1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * fs/f2fs/super.c 4 * 5 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 6 * http://www.samsung.com/ 7 */ 8 #include <linux/module.h> 9 #include <linux/init.h> 10 #include <linux/fs.h> 11 #include <linux/fs_context.h> 12 #include <linux/sched/mm.h> 13 #include <linux/statfs.h> 14 #include <linux/buffer_head.h> 15 #include <linux/kthread.h> 16 #include <linux/parser.h> 17 #include <linux/mount.h> 18 #include <linux/seq_file.h> 19 #include <linux/proc_fs.h> 20 #include <linux/random.h> 21 #include <linux/exportfs.h> 22 #include <linux/blkdev.h> 23 #include <linux/quotaops.h> 24 #include <linux/f2fs_fs.h> 25 #include <linux/sysfs.h> 26 #include <linux/quota.h> 27 #include <linux/unicode.h> 28 #include <linux/part_stat.h> 29 #include <linux/zstd.h> 30 #include <linux/lz4.h> 31 32 #include "f2fs.h" 33 #include "node.h" 34 #include "segment.h" 35 #include "xattr.h" 36 #include "gc.h" 37 #include "iostat.h" 38 39 #define CREATE_TRACE_POINTS 40 #include <trace/events/f2fs.h> 41 42 static struct kmem_cache *f2fs_inode_cachep; 43 44 #ifdef CONFIG_F2FS_FAULT_INJECTION 45 46 const char *f2fs_fault_name[FAULT_MAX] = { 47 [FAULT_KMALLOC] = "kmalloc", 48 [FAULT_KVMALLOC] = "kvmalloc", 49 [FAULT_PAGE_ALLOC] = "page alloc", 50 [FAULT_PAGE_GET] = "page get", 51 [FAULT_ALLOC_NID] = "alloc nid", 52 [FAULT_ORPHAN] = "orphan", 53 [FAULT_BLOCK] = "no more block", 54 [FAULT_DIR_DEPTH] = "too big dir depth", 55 [FAULT_EVICT_INODE] = "evict_inode fail", 56 [FAULT_TRUNCATE] = "truncate fail", 57 [FAULT_READ_IO] = "read IO error", 58 [FAULT_CHECKPOINT] = "checkpoint error", 59 [FAULT_DISCARD] = "discard error", 60 [FAULT_WRITE_IO] = "write IO error", 61 [FAULT_SLAB_ALLOC] = "slab alloc", 62 [FAULT_DQUOT_INIT] = "dquot initialize", 63 [FAULT_LOCK_OP] = "lock_op", 64 [FAULT_BLKADDR] = "invalid blkaddr", 65 }; 66 67 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate, 68 unsigned int type) 69 { 70 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info; 71 72 if (rate) { 73 atomic_set(&ffi->inject_ops, 0); 74 ffi->inject_rate = rate; 75 } 76 77 if (type) 78 ffi->inject_type = type; 79 80 if (!rate && !type) 81 memset(ffi, 0, sizeof(struct f2fs_fault_info)); 82 } 83 #endif 84 85 /* f2fs-wide shrinker description */ 86 static struct shrinker f2fs_shrinker_info = { 87 .scan_objects = f2fs_shrink_scan, 88 .count_objects = f2fs_shrink_count, 89 .seeks = DEFAULT_SEEKS, 90 }; 91 92 enum { 93 Opt_gc_background, 94 Opt_disable_roll_forward, 95 Opt_norecovery, 96 Opt_discard, 97 Opt_nodiscard, 98 Opt_noheap, 99 Opt_heap, 100 Opt_user_xattr, 101 Opt_nouser_xattr, 102 Opt_acl, 103 Opt_noacl, 104 Opt_active_logs, 105 Opt_disable_ext_identify, 106 Opt_inline_xattr, 107 Opt_noinline_xattr, 108 Opt_inline_xattr_size, 109 Opt_inline_data, 110 Opt_inline_dentry, 111 Opt_noinline_dentry, 112 Opt_flush_merge, 113 Opt_noflush_merge, 114 Opt_barrier, 115 Opt_nobarrier, 116 Opt_fastboot, 117 Opt_extent_cache, 118 Opt_noextent_cache, 119 Opt_noinline_data, 120 Opt_data_flush, 121 Opt_reserve_root, 122 Opt_resgid, 123 Opt_resuid, 124 Opt_mode, 125 Opt_io_size_bits, 126 Opt_fault_injection, 127 Opt_fault_type, 128 Opt_lazytime, 129 Opt_nolazytime, 130 Opt_quota, 131 Opt_noquota, 132 Opt_usrquota, 133 Opt_grpquota, 134 Opt_prjquota, 135 Opt_usrjquota, 136 Opt_grpjquota, 137 Opt_prjjquota, 138 Opt_offusrjquota, 139 Opt_offgrpjquota, 140 Opt_offprjjquota, 141 Opt_jqfmt_vfsold, 142 Opt_jqfmt_vfsv0, 143 Opt_jqfmt_vfsv1, 144 Opt_alloc, 145 Opt_fsync, 146 Opt_test_dummy_encryption, 147 Opt_inlinecrypt, 148 Opt_checkpoint_disable, 149 Opt_checkpoint_disable_cap, 150 Opt_checkpoint_disable_cap_perc, 151 Opt_checkpoint_enable, 152 Opt_checkpoint_merge, 153 Opt_nocheckpoint_merge, 154 Opt_compress_algorithm, 155 Opt_compress_log_size, 156 Opt_compress_extension, 157 Opt_nocompress_extension, 158 Opt_compress_chksum, 159 Opt_compress_mode, 160 Opt_compress_cache, 161 Opt_atgc, 162 Opt_gc_merge, 163 Opt_nogc_merge, 164 Opt_discard_unit, 165 Opt_memory_mode, 166 Opt_age_extent_cache, 167 Opt_errors, 168 Opt_err, 169 }; 170 171 static match_table_t f2fs_tokens = { 172 {Opt_gc_background, "background_gc=%s"}, 173 {Opt_disable_roll_forward, "disable_roll_forward"}, 174 {Opt_norecovery, "norecovery"}, 175 {Opt_discard, "discard"}, 176 {Opt_nodiscard, "nodiscard"}, 177 {Opt_noheap, "no_heap"}, 178 {Opt_heap, "heap"}, 179 {Opt_user_xattr, "user_xattr"}, 180 {Opt_nouser_xattr, "nouser_xattr"}, 181 {Opt_acl, "acl"}, 182 {Opt_noacl, "noacl"}, 183 {Opt_active_logs, "active_logs=%u"}, 184 {Opt_disable_ext_identify, "disable_ext_identify"}, 185 {Opt_inline_xattr, "inline_xattr"}, 186 {Opt_noinline_xattr, "noinline_xattr"}, 187 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 188 {Opt_inline_data, "inline_data"}, 189 {Opt_inline_dentry, "inline_dentry"}, 190 {Opt_noinline_dentry, "noinline_dentry"}, 191 {Opt_flush_merge, "flush_merge"}, 192 {Opt_noflush_merge, "noflush_merge"}, 193 {Opt_barrier, "barrier"}, 194 {Opt_nobarrier, "nobarrier"}, 195 {Opt_fastboot, "fastboot"}, 196 {Opt_extent_cache, "extent_cache"}, 197 {Opt_noextent_cache, "noextent_cache"}, 198 {Opt_noinline_data, "noinline_data"}, 199 {Opt_data_flush, "data_flush"}, 200 {Opt_reserve_root, "reserve_root=%u"}, 201 {Opt_resgid, "resgid=%u"}, 202 {Opt_resuid, "resuid=%u"}, 203 {Opt_mode, "mode=%s"}, 204 {Opt_io_size_bits, "io_bits=%u"}, 205 {Opt_fault_injection, "fault_injection=%u"}, 206 {Opt_fault_type, "fault_type=%u"}, 207 {Opt_lazytime, "lazytime"}, 208 {Opt_nolazytime, "nolazytime"}, 209 {Opt_quota, "quota"}, 210 {Opt_noquota, "noquota"}, 211 {Opt_usrquota, "usrquota"}, 212 {Opt_grpquota, "grpquota"}, 213 {Opt_prjquota, "prjquota"}, 214 {Opt_usrjquota, "usrjquota=%s"}, 215 {Opt_grpjquota, "grpjquota=%s"}, 216 {Opt_prjjquota, "prjjquota=%s"}, 217 {Opt_offusrjquota, "usrjquota="}, 218 {Opt_offgrpjquota, "grpjquota="}, 219 {Opt_offprjjquota, "prjjquota="}, 220 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 221 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 222 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 223 {Opt_alloc, "alloc_mode=%s"}, 224 {Opt_fsync, "fsync_mode=%s"}, 225 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 226 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 227 {Opt_inlinecrypt, "inlinecrypt"}, 228 {Opt_checkpoint_disable, "checkpoint=disable"}, 229 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 230 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 231 {Opt_checkpoint_enable, "checkpoint=enable"}, 232 {Opt_checkpoint_merge, "checkpoint_merge"}, 233 {Opt_nocheckpoint_merge, "nocheckpoint_merge"}, 234 {Opt_compress_algorithm, "compress_algorithm=%s"}, 235 {Opt_compress_log_size, "compress_log_size=%u"}, 236 {Opt_compress_extension, "compress_extension=%s"}, 237 {Opt_nocompress_extension, "nocompress_extension=%s"}, 238 {Opt_compress_chksum, "compress_chksum"}, 239 {Opt_compress_mode, "compress_mode=%s"}, 240 {Opt_compress_cache, "compress_cache"}, 241 {Opt_atgc, "atgc"}, 242 {Opt_gc_merge, "gc_merge"}, 243 {Opt_nogc_merge, "nogc_merge"}, 244 {Opt_discard_unit, "discard_unit=%s"}, 245 {Opt_memory_mode, "memory=%s"}, 246 {Opt_age_extent_cache, "age_extent_cache"}, 247 {Opt_errors, "errors=%s"}, 248 {Opt_err, NULL}, 249 }; 250 251 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) 252 { 253 struct va_format vaf; 254 va_list args; 255 int level; 256 257 va_start(args, fmt); 258 259 level = printk_get_level(fmt); 260 vaf.fmt = printk_skip_level(fmt); 261 vaf.va = &args; 262 printk("%c%cF2FS-fs (%s): %pV\n", 263 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 264 265 va_end(args); 266 } 267 268 #if IS_ENABLED(CONFIG_UNICODE) 269 static const struct f2fs_sb_encodings { 270 __u16 magic; 271 char *name; 272 unsigned int version; 273 } f2fs_sb_encoding_map[] = { 274 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 275 }; 276 277 static const struct f2fs_sb_encodings * 278 f2fs_sb_read_encoding(const struct f2fs_super_block *sb) 279 { 280 __u16 magic = le16_to_cpu(sb->s_encoding); 281 int i; 282 283 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 284 if (magic == f2fs_sb_encoding_map[i].magic) 285 return &f2fs_sb_encoding_map[i]; 286 287 return NULL; 288 } 289 290 struct kmem_cache *f2fs_cf_name_slab; 291 static int __init f2fs_create_casefold_cache(void) 292 { 293 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name", 294 F2FS_NAME_LEN); 295 return f2fs_cf_name_slab ? 0 : -ENOMEM; 296 } 297 298 static void f2fs_destroy_casefold_cache(void) 299 { 300 kmem_cache_destroy(f2fs_cf_name_slab); 301 } 302 #else 303 static int __init f2fs_create_casefold_cache(void) { return 0; } 304 static void f2fs_destroy_casefold_cache(void) { } 305 #endif 306 307 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 308 { 309 block_t limit = min((sbi->user_block_count >> 3), 310 sbi->user_block_count - sbi->reserved_blocks); 311 312 /* limit is 12.5% */ 313 if (test_opt(sbi, RESERVE_ROOT) && 314 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 315 F2FS_OPTION(sbi).root_reserved_blocks = limit; 316 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 317 F2FS_OPTION(sbi).root_reserved_blocks); 318 } 319 if (!test_opt(sbi, RESERVE_ROOT) && 320 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 321 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 322 !gid_eq(F2FS_OPTION(sbi).s_resgid, 323 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 324 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 325 from_kuid_munged(&init_user_ns, 326 F2FS_OPTION(sbi).s_resuid), 327 from_kgid_munged(&init_user_ns, 328 F2FS_OPTION(sbi).s_resgid)); 329 } 330 331 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi) 332 { 333 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec; 334 unsigned int avg_vblocks; 335 unsigned int wanted_reserved_segments; 336 block_t avail_user_block_count; 337 338 if (!F2FS_IO_ALIGNED(sbi)) 339 return 0; 340 341 /* average valid block count in section in worst case */ 342 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi); 343 344 /* 345 * we need enough free space when migrating one section in worst case 346 */ 347 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) * 348 reserved_segments(sbi); 349 wanted_reserved_segments -= reserved_segments(sbi); 350 351 avail_user_block_count = sbi->user_block_count - 352 sbi->current_reserved_blocks - 353 F2FS_OPTION(sbi).root_reserved_blocks; 354 355 if (wanted_reserved_segments * sbi->blocks_per_seg > 356 avail_user_block_count) { 357 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u", 358 wanted_reserved_segments, 359 avail_user_block_count >> sbi->log_blocks_per_seg); 360 return -ENOSPC; 361 } 362 363 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments; 364 365 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u", 366 wanted_reserved_segments); 367 368 return 0; 369 } 370 371 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 372 { 373 if (!F2FS_OPTION(sbi).unusable_cap_perc) 374 return; 375 376 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 377 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 378 else 379 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 380 F2FS_OPTION(sbi).unusable_cap_perc; 381 382 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 383 F2FS_OPTION(sbi).unusable_cap, 384 F2FS_OPTION(sbi).unusable_cap_perc); 385 } 386 387 static void init_once(void *foo) 388 { 389 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 390 391 inode_init_once(&fi->vfs_inode); 392 } 393 394 #ifdef CONFIG_QUOTA 395 static const char * const quotatypes[] = INITQFNAMES; 396 #define QTYPE2NAME(t) (quotatypes[t]) 397 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 398 substring_t *args) 399 { 400 struct f2fs_sb_info *sbi = F2FS_SB(sb); 401 char *qname; 402 int ret = -EINVAL; 403 404 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 405 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 406 return -EINVAL; 407 } 408 if (f2fs_sb_has_quota_ino(sbi)) { 409 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 410 return 0; 411 } 412 413 qname = match_strdup(args); 414 if (!qname) { 415 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 416 return -ENOMEM; 417 } 418 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 419 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 420 ret = 0; 421 else 422 f2fs_err(sbi, "%s quota file already specified", 423 QTYPE2NAME(qtype)); 424 goto errout; 425 } 426 if (strchr(qname, '/')) { 427 f2fs_err(sbi, "quotafile must be on filesystem root"); 428 goto errout; 429 } 430 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 431 set_opt(sbi, QUOTA); 432 return 0; 433 errout: 434 kfree(qname); 435 return ret; 436 } 437 438 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 439 { 440 struct f2fs_sb_info *sbi = F2FS_SB(sb); 441 442 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 443 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 444 return -EINVAL; 445 } 446 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 447 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 448 return 0; 449 } 450 451 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 452 { 453 /* 454 * We do the test below only for project quotas. 'usrquota' and 455 * 'grpquota' mount options are allowed even without quota feature 456 * to support legacy quotas in quota files. 457 */ 458 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 459 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 460 return -1; 461 } 462 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 463 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 464 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 465 if (test_opt(sbi, USRQUOTA) && 466 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 467 clear_opt(sbi, USRQUOTA); 468 469 if (test_opt(sbi, GRPQUOTA) && 470 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 471 clear_opt(sbi, GRPQUOTA); 472 473 if (test_opt(sbi, PRJQUOTA) && 474 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 475 clear_opt(sbi, PRJQUOTA); 476 477 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 478 test_opt(sbi, PRJQUOTA)) { 479 f2fs_err(sbi, "old and new quota format mixing"); 480 return -1; 481 } 482 483 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 484 f2fs_err(sbi, "journaled quota format not specified"); 485 return -1; 486 } 487 } 488 489 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 490 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 491 F2FS_OPTION(sbi).s_jquota_fmt = 0; 492 } 493 return 0; 494 } 495 #endif 496 497 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 498 const char *opt, 499 const substring_t *arg, 500 bool is_remount) 501 { 502 struct f2fs_sb_info *sbi = F2FS_SB(sb); 503 struct fs_parameter param = { 504 .type = fs_value_is_string, 505 .string = arg->from ? arg->from : "", 506 }; 507 struct fscrypt_dummy_policy *policy = 508 &F2FS_OPTION(sbi).dummy_enc_policy; 509 int err; 510 511 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) { 512 f2fs_warn(sbi, "test_dummy_encryption option not supported"); 513 return -EINVAL; 514 } 515 516 if (!f2fs_sb_has_encrypt(sbi)) { 517 f2fs_err(sbi, "Encrypt feature is off"); 518 return -EINVAL; 519 } 520 521 /* 522 * This mount option is just for testing, and it's not worthwhile to 523 * implement the extra complexity (e.g. RCU protection) that would be 524 * needed to allow it to be set or changed during remount. We do allow 525 * it to be specified during remount, but only if there is no change. 526 */ 527 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) { 528 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 529 return -EINVAL; 530 } 531 532 err = fscrypt_parse_test_dummy_encryption(¶m, policy); 533 if (err) { 534 if (err == -EEXIST) 535 f2fs_warn(sbi, 536 "Can't change test_dummy_encryption on remount"); 537 else if (err == -EINVAL) 538 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 539 opt); 540 else 541 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 542 opt, err); 543 return -EINVAL; 544 } 545 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 546 return 0; 547 } 548 549 #ifdef CONFIG_F2FS_FS_COMPRESSION 550 /* 551 * 1. The same extension name cannot not appear in both compress and non-compress extension 552 * at the same time. 553 * 2. If the compress extension specifies all files, the types specified by the non-compress 554 * extension will be treated as special cases and will not be compressed. 555 * 3. Don't allow the non-compress extension specifies all files. 556 */ 557 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi) 558 { 559 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 560 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 561 int ext_cnt, noext_cnt, index = 0, no_index = 0; 562 563 ext = F2FS_OPTION(sbi).extensions; 564 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 565 noext = F2FS_OPTION(sbi).noextensions; 566 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 567 568 if (!noext_cnt) 569 return 0; 570 571 for (no_index = 0; no_index < noext_cnt; no_index++) { 572 if (!strcasecmp("*", noext[no_index])) { 573 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files"); 574 return -EINVAL; 575 } 576 for (index = 0; index < ext_cnt; index++) { 577 if (!strcasecmp(ext[index], noext[no_index])) { 578 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension", 579 ext[index]); 580 return -EINVAL; 581 } 582 } 583 } 584 return 0; 585 } 586 587 #ifdef CONFIG_F2FS_FS_LZ4 588 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str) 589 { 590 #ifdef CONFIG_F2FS_FS_LZ4HC 591 unsigned int level; 592 593 if (strlen(str) == 3) { 594 F2FS_OPTION(sbi).compress_level = LZ4HC_DEFAULT_CLEVEL; 595 return 0; 596 } 597 598 str += 3; 599 600 if (str[0] != ':') { 601 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 602 return -EINVAL; 603 } 604 if (kstrtouint(str + 1, 10, &level)) 605 return -EINVAL; 606 607 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) { 608 f2fs_info(sbi, "invalid lz4hc compress level: %d", level); 609 return -EINVAL; 610 } 611 612 F2FS_OPTION(sbi).compress_level = level; 613 return 0; 614 #else 615 if (strlen(str) == 3) { 616 F2FS_OPTION(sbi).compress_level = 0; 617 return 0; 618 } 619 f2fs_info(sbi, "kernel doesn't support lz4hc compression"); 620 return -EINVAL; 621 #endif 622 } 623 #endif 624 625 #ifdef CONFIG_F2FS_FS_ZSTD 626 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str) 627 { 628 unsigned int level; 629 int len = 4; 630 631 if (strlen(str) == len) { 632 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 633 return 0; 634 } 635 636 str += len; 637 638 if (str[0] != ':') { 639 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 640 return -EINVAL; 641 } 642 if (kstrtouint(str + 1, 10, &level)) 643 return -EINVAL; 644 645 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) { 646 f2fs_info(sbi, "invalid zstd compress level: %d", level); 647 return -EINVAL; 648 } 649 650 F2FS_OPTION(sbi).compress_level = level; 651 return 0; 652 } 653 #endif 654 #endif 655 656 static int parse_options(struct super_block *sb, char *options, bool is_remount) 657 { 658 struct f2fs_sb_info *sbi = F2FS_SB(sb); 659 substring_t args[MAX_OPT_ARGS]; 660 #ifdef CONFIG_F2FS_FS_COMPRESSION 661 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 662 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 663 int ext_cnt, noext_cnt; 664 #endif 665 char *p, *name; 666 int arg = 0; 667 kuid_t uid; 668 kgid_t gid; 669 int ret; 670 671 if (!options) 672 goto default_check; 673 674 while ((p = strsep(&options, ",")) != NULL) { 675 int token; 676 677 if (!*p) 678 continue; 679 /* 680 * Initialize args struct so we know whether arg was 681 * found; some options take optional arguments. 682 */ 683 args[0].to = args[0].from = NULL; 684 token = match_token(p, f2fs_tokens, args); 685 686 switch (token) { 687 case Opt_gc_background: 688 name = match_strdup(&args[0]); 689 690 if (!name) 691 return -ENOMEM; 692 if (!strcmp(name, "on")) { 693 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 694 } else if (!strcmp(name, "off")) { 695 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 696 } else if (!strcmp(name, "sync")) { 697 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 698 } else { 699 kfree(name); 700 return -EINVAL; 701 } 702 kfree(name); 703 break; 704 case Opt_disable_roll_forward: 705 set_opt(sbi, DISABLE_ROLL_FORWARD); 706 break; 707 case Opt_norecovery: 708 /* this option mounts f2fs with ro */ 709 set_opt(sbi, NORECOVERY); 710 if (!f2fs_readonly(sb)) 711 return -EINVAL; 712 break; 713 case Opt_discard: 714 if (!f2fs_hw_support_discard(sbi)) { 715 f2fs_warn(sbi, "device does not support discard"); 716 break; 717 } 718 set_opt(sbi, DISCARD); 719 break; 720 case Opt_nodiscard: 721 if (f2fs_hw_should_discard(sbi)) { 722 f2fs_warn(sbi, "discard is required for zoned block devices"); 723 return -EINVAL; 724 } 725 clear_opt(sbi, DISCARD); 726 break; 727 case Opt_noheap: 728 set_opt(sbi, NOHEAP); 729 break; 730 case Opt_heap: 731 clear_opt(sbi, NOHEAP); 732 break; 733 #ifdef CONFIG_F2FS_FS_XATTR 734 case Opt_user_xattr: 735 set_opt(sbi, XATTR_USER); 736 break; 737 case Opt_nouser_xattr: 738 clear_opt(sbi, XATTR_USER); 739 break; 740 case Opt_inline_xattr: 741 set_opt(sbi, INLINE_XATTR); 742 break; 743 case Opt_noinline_xattr: 744 clear_opt(sbi, INLINE_XATTR); 745 break; 746 case Opt_inline_xattr_size: 747 if (args->from && match_int(args, &arg)) 748 return -EINVAL; 749 set_opt(sbi, INLINE_XATTR_SIZE); 750 F2FS_OPTION(sbi).inline_xattr_size = arg; 751 break; 752 #else 753 case Opt_user_xattr: 754 f2fs_info(sbi, "user_xattr options not supported"); 755 break; 756 case Opt_nouser_xattr: 757 f2fs_info(sbi, "nouser_xattr options not supported"); 758 break; 759 case Opt_inline_xattr: 760 f2fs_info(sbi, "inline_xattr options not supported"); 761 break; 762 case Opt_noinline_xattr: 763 f2fs_info(sbi, "noinline_xattr options not supported"); 764 break; 765 #endif 766 #ifdef CONFIG_F2FS_FS_POSIX_ACL 767 case Opt_acl: 768 set_opt(sbi, POSIX_ACL); 769 break; 770 case Opt_noacl: 771 clear_opt(sbi, POSIX_ACL); 772 break; 773 #else 774 case Opt_acl: 775 f2fs_info(sbi, "acl options not supported"); 776 break; 777 case Opt_noacl: 778 f2fs_info(sbi, "noacl options not supported"); 779 break; 780 #endif 781 case Opt_active_logs: 782 if (args->from && match_int(args, &arg)) 783 return -EINVAL; 784 if (arg != 2 && arg != 4 && 785 arg != NR_CURSEG_PERSIST_TYPE) 786 return -EINVAL; 787 F2FS_OPTION(sbi).active_logs = arg; 788 break; 789 case Opt_disable_ext_identify: 790 set_opt(sbi, DISABLE_EXT_IDENTIFY); 791 break; 792 case Opt_inline_data: 793 set_opt(sbi, INLINE_DATA); 794 break; 795 case Opt_inline_dentry: 796 set_opt(sbi, INLINE_DENTRY); 797 break; 798 case Opt_noinline_dentry: 799 clear_opt(sbi, INLINE_DENTRY); 800 break; 801 case Opt_flush_merge: 802 set_opt(sbi, FLUSH_MERGE); 803 break; 804 case Opt_noflush_merge: 805 clear_opt(sbi, FLUSH_MERGE); 806 break; 807 case Opt_nobarrier: 808 set_opt(sbi, NOBARRIER); 809 break; 810 case Opt_barrier: 811 clear_opt(sbi, NOBARRIER); 812 break; 813 case Opt_fastboot: 814 set_opt(sbi, FASTBOOT); 815 break; 816 case Opt_extent_cache: 817 set_opt(sbi, READ_EXTENT_CACHE); 818 break; 819 case Opt_noextent_cache: 820 clear_opt(sbi, READ_EXTENT_CACHE); 821 break; 822 case Opt_noinline_data: 823 clear_opt(sbi, INLINE_DATA); 824 break; 825 case Opt_data_flush: 826 set_opt(sbi, DATA_FLUSH); 827 break; 828 case Opt_reserve_root: 829 if (args->from && match_int(args, &arg)) 830 return -EINVAL; 831 if (test_opt(sbi, RESERVE_ROOT)) { 832 f2fs_info(sbi, "Preserve previous reserve_root=%u", 833 F2FS_OPTION(sbi).root_reserved_blocks); 834 } else { 835 F2FS_OPTION(sbi).root_reserved_blocks = arg; 836 set_opt(sbi, RESERVE_ROOT); 837 } 838 break; 839 case Opt_resuid: 840 if (args->from && match_int(args, &arg)) 841 return -EINVAL; 842 uid = make_kuid(current_user_ns(), arg); 843 if (!uid_valid(uid)) { 844 f2fs_err(sbi, "Invalid uid value %d", arg); 845 return -EINVAL; 846 } 847 F2FS_OPTION(sbi).s_resuid = uid; 848 break; 849 case Opt_resgid: 850 if (args->from && match_int(args, &arg)) 851 return -EINVAL; 852 gid = make_kgid(current_user_ns(), arg); 853 if (!gid_valid(gid)) { 854 f2fs_err(sbi, "Invalid gid value %d", arg); 855 return -EINVAL; 856 } 857 F2FS_OPTION(sbi).s_resgid = gid; 858 break; 859 case Opt_mode: 860 name = match_strdup(&args[0]); 861 862 if (!name) 863 return -ENOMEM; 864 if (!strcmp(name, "adaptive")) { 865 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 866 } else if (!strcmp(name, "lfs")) { 867 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 868 } else if (!strcmp(name, "fragment:segment")) { 869 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG; 870 } else if (!strcmp(name, "fragment:block")) { 871 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK; 872 } else { 873 kfree(name); 874 return -EINVAL; 875 } 876 kfree(name); 877 break; 878 case Opt_io_size_bits: 879 if (args->from && match_int(args, &arg)) 880 return -EINVAL; 881 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) { 882 f2fs_warn(sbi, "Not support %ld, larger than %d", 883 BIT(arg), BIO_MAX_VECS); 884 return -EINVAL; 885 } 886 F2FS_OPTION(sbi).write_io_size_bits = arg; 887 break; 888 #ifdef CONFIG_F2FS_FAULT_INJECTION 889 case Opt_fault_injection: 890 if (args->from && match_int(args, &arg)) 891 return -EINVAL; 892 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 893 set_opt(sbi, FAULT_INJECTION); 894 break; 895 896 case Opt_fault_type: 897 if (args->from && match_int(args, &arg)) 898 return -EINVAL; 899 f2fs_build_fault_attr(sbi, 0, arg); 900 set_opt(sbi, FAULT_INJECTION); 901 break; 902 #else 903 case Opt_fault_injection: 904 f2fs_info(sbi, "fault_injection options not supported"); 905 break; 906 907 case Opt_fault_type: 908 f2fs_info(sbi, "fault_type options not supported"); 909 break; 910 #endif 911 case Opt_lazytime: 912 sb->s_flags |= SB_LAZYTIME; 913 break; 914 case Opt_nolazytime: 915 sb->s_flags &= ~SB_LAZYTIME; 916 break; 917 #ifdef CONFIG_QUOTA 918 case Opt_quota: 919 case Opt_usrquota: 920 set_opt(sbi, USRQUOTA); 921 break; 922 case Opt_grpquota: 923 set_opt(sbi, GRPQUOTA); 924 break; 925 case Opt_prjquota: 926 set_opt(sbi, PRJQUOTA); 927 break; 928 case Opt_usrjquota: 929 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 930 if (ret) 931 return ret; 932 break; 933 case Opt_grpjquota: 934 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 935 if (ret) 936 return ret; 937 break; 938 case Opt_prjjquota: 939 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 940 if (ret) 941 return ret; 942 break; 943 case Opt_offusrjquota: 944 ret = f2fs_clear_qf_name(sb, USRQUOTA); 945 if (ret) 946 return ret; 947 break; 948 case Opt_offgrpjquota: 949 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 950 if (ret) 951 return ret; 952 break; 953 case Opt_offprjjquota: 954 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 955 if (ret) 956 return ret; 957 break; 958 case Opt_jqfmt_vfsold: 959 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 960 break; 961 case Opt_jqfmt_vfsv0: 962 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 963 break; 964 case Opt_jqfmt_vfsv1: 965 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 966 break; 967 case Opt_noquota: 968 clear_opt(sbi, QUOTA); 969 clear_opt(sbi, USRQUOTA); 970 clear_opt(sbi, GRPQUOTA); 971 clear_opt(sbi, PRJQUOTA); 972 break; 973 #else 974 case Opt_quota: 975 case Opt_usrquota: 976 case Opt_grpquota: 977 case Opt_prjquota: 978 case Opt_usrjquota: 979 case Opt_grpjquota: 980 case Opt_prjjquota: 981 case Opt_offusrjquota: 982 case Opt_offgrpjquota: 983 case Opt_offprjjquota: 984 case Opt_jqfmt_vfsold: 985 case Opt_jqfmt_vfsv0: 986 case Opt_jqfmt_vfsv1: 987 case Opt_noquota: 988 f2fs_info(sbi, "quota operations not supported"); 989 break; 990 #endif 991 case Opt_alloc: 992 name = match_strdup(&args[0]); 993 if (!name) 994 return -ENOMEM; 995 996 if (!strcmp(name, "default")) { 997 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 998 } else if (!strcmp(name, "reuse")) { 999 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 1000 } else { 1001 kfree(name); 1002 return -EINVAL; 1003 } 1004 kfree(name); 1005 break; 1006 case Opt_fsync: 1007 name = match_strdup(&args[0]); 1008 if (!name) 1009 return -ENOMEM; 1010 if (!strcmp(name, "posix")) { 1011 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1012 } else if (!strcmp(name, "strict")) { 1013 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 1014 } else if (!strcmp(name, "nobarrier")) { 1015 F2FS_OPTION(sbi).fsync_mode = 1016 FSYNC_MODE_NOBARRIER; 1017 } else { 1018 kfree(name); 1019 return -EINVAL; 1020 } 1021 kfree(name); 1022 break; 1023 case Opt_test_dummy_encryption: 1024 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 1025 is_remount); 1026 if (ret) 1027 return ret; 1028 break; 1029 case Opt_inlinecrypt: 1030 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 1031 sb->s_flags |= SB_INLINECRYPT; 1032 #else 1033 f2fs_info(sbi, "inline encryption not supported"); 1034 #endif 1035 break; 1036 case Opt_checkpoint_disable_cap_perc: 1037 if (args->from && match_int(args, &arg)) 1038 return -EINVAL; 1039 if (arg < 0 || arg > 100) 1040 return -EINVAL; 1041 F2FS_OPTION(sbi).unusable_cap_perc = arg; 1042 set_opt(sbi, DISABLE_CHECKPOINT); 1043 break; 1044 case Opt_checkpoint_disable_cap: 1045 if (args->from && match_int(args, &arg)) 1046 return -EINVAL; 1047 F2FS_OPTION(sbi).unusable_cap = arg; 1048 set_opt(sbi, DISABLE_CHECKPOINT); 1049 break; 1050 case Opt_checkpoint_disable: 1051 set_opt(sbi, DISABLE_CHECKPOINT); 1052 break; 1053 case Opt_checkpoint_enable: 1054 clear_opt(sbi, DISABLE_CHECKPOINT); 1055 break; 1056 case Opt_checkpoint_merge: 1057 set_opt(sbi, MERGE_CHECKPOINT); 1058 break; 1059 case Opt_nocheckpoint_merge: 1060 clear_opt(sbi, MERGE_CHECKPOINT); 1061 break; 1062 #ifdef CONFIG_F2FS_FS_COMPRESSION 1063 case Opt_compress_algorithm: 1064 if (!f2fs_sb_has_compression(sbi)) { 1065 f2fs_info(sbi, "Image doesn't support compression"); 1066 break; 1067 } 1068 name = match_strdup(&args[0]); 1069 if (!name) 1070 return -ENOMEM; 1071 if (!strcmp(name, "lzo")) { 1072 #ifdef CONFIG_F2FS_FS_LZO 1073 F2FS_OPTION(sbi).compress_level = 0; 1074 F2FS_OPTION(sbi).compress_algorithm = 1075 COMPRESS_LZO; 1076 #else 1077 f2fs_info(sbi, "kernel doesn't support lzo compression"); 1078 #endif 1079 } else if (!strncmp(name, "lz4", 3)) { 1080 #ifdef CONFIG_F2FS_FS_LZ4 1081 ret = f2fs_set_lz4hc_level(sbi, name); 1082 if (ret) { 1083 kfree(name); 1084 return -EINVAL; 1085 } 1086 F2FS_OPTION(sbi).compress_algorithm = 1087 COMPRESS_LZ4; 1088 #else 1089 f2fs_info(sbi, "kernel doesn't support lz4 compression"); 1090 #endif 1091 } else if (!strncmp(name, "zstd", 4)) { 1092 #ifdef CONFIG_F2FS_FS_ZSTD 1093 ret = f2fs_set_zstd_level(sbi, name); 1094 if (ret) { 1095 kfree(name); 1096 return -EINVAL; 1097 } 1098 F2FS_OPTION(sbi).compress_algorithm = 1099 COMPRESS_ZSTD; 1100 #else 1101 f2fs_info(sbi, "kernel doesn't support zstd compression"); 1102 #endif 1103 } else if (!strcmp(name, "lzo-rle")) { 1104 #ifdef CONFIG_F2FS_FS_LZORLE 1105 F2FS_OPTION(sbi).compress_level = 0; 1106 F2FS_OPTION(sbi).compress_algorithm = 1107 COMPRESS_LZORLE; 1108 #else 1109 f2fs_info(sbi, "kernel doesn't support lzorle compression"); 1110 #endif 1111 } else { 1112 kfree(name); 1113 return -EINVAL; 1114 } 1115 kfree(name); 1116 break; 1117 case Opt_compress_log_size: 1118 if (!f2fs_sb_has_compression(sbi)) { 1119 f2fs_info(sbi, "Image doesn't support compression"); 1120 break; 1121 } 1122 if (args->from && match_int(args, &arg)) 1123 return -EINVAL; 1124 if (arg < MIN_COMPRESS_LOG_SIZE || 1125 arg > MAX_COMPRESS_LOG_SIZE) { 1126 f2fs_err(sbi, 1127 "Compress cluster log size is out of range"); 1128 return -EINVAL; 1129 } 1130 F2FS_OPTION(sbi).compress_log_size = arg; 1131 break; 1132 case Opt_compress_extension: 1133 if (!f2fs_sb_has_compression(sbi)) { 1134 f2fs_info(sbi, "Image doesn't support compression"); 1135 break; 1136 } 1137 name = match_strdup(&args[0]); 1138 if (!name) 1139 return -ENOMEM; 1140 1141 ext = F2FS_OPTION(sbi).extensions; 1142 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 1143 1144 if (strlen(name) >= F2FS_EXTENSION_LEN || 1145 ext_cnt >= COMPRESS_EXT_NUM) { 1146 f2fs_err(sbi, 1147 "invalid extension length/number"); 1148 kfree(name); 1149 return -EINVAL; 1150 } 1151 1152 strcpy(ext[ext_cnt], name); 1153 F2FS_OPTION(sbi).compress_ext_cnt++; 1154 kfree(name); 1155 break; 1156 case Opt_nocompress_extension: 1157 if (!f2fs_sb_has_compression(sbi)) { 1158 f2fs_info(sbi, "Image doesn't support compression"); 1159 break; 1160 } 1161 name = match_strdup(&args[0]); 1162 if (!name) 1163 return -ENOMEM; 1164 1165 noext = F2FS_OPTION(sbi).noextensions; 1166 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 1167 1168 if (strlen(name) >= F2FS_EXTENSION_LEN || 1169 noext_cnt >= COMPRESS_EXT_NUM) { 1170 f2fs_err(sbi, 1171 "invalid extension length/number"); 1172 kfree(name); 1173 return -EINVAL; 1174 } 1175 1176 strcpy(noext[noext_cnt], name); 1177 F2FS_OPTION(sbi).nocompress_ext_cnt++; 1178 kfree(name); 1179 break; 1180 case Opt_compress_chksum: 1181 if (!f2fs_sb_has_compression(sbi)) { 1182 f2fs_info(sbi, "Image doesn't support compression"); 1183 break; 1184 } 1185 F2FS_OPTION(sbi).compress_chksum = true; 1186 break; 1187 case Opt_compress_mode: 1188 if (!f2fs_sb_has_compression(sbi)) { 1189 f2fs_info(sbi, "Image doesn't support compression"); 1190 break; 1191 } 1192 name = match_strdup(&args[0]); 1193 if (!name) 1194 return -ENOMEM; 1195 if (!strcmp(name, "fs")) { 1196 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 1197 } else if (!strcmp(name, "user")) { 1198 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER; 1199 } else { 1200 kfree(name); 1201 return -EINVAL; 1202 } 1203 kfree(name); 1204 break; 1205 case Opt_compress_cache: 1206 if (!f2fs_sb_has_compression(sbi)) { 1207 f2fs_info(sbi, "Image doesn't support compression"); 1208 break; 1209 } 1210 set_opt(sbi, COMPRESS_CACHE); 1211 break; 1212 #else 1213 case Opt_compress_algorithm: 1214 case Opt_compress_log_size: 1215 case Opt_compress_extension: 1216 case Opt_nocompress_extension: 1217 case Opt_compress_chksum: 1218 case Opt_compress_mode: 1219 case Opt_compress_cache: 1220 f2fs_info(sbi, "compression options not supported"); 1221 break; 1222 #endif 1223 case Opt_atgc: 1224 set_opt(sbi, ATGC); 1225 break; 1226 case Opt_gc_merge: 1227 set_opt(sbi, GC_MERGE); 1228 break; 1229 case Opt_nogc_merge: 1230 clear_opt(sbi, GC_MERGE); 1231 break; 1232 case Opt_discard_unit: 1233 name = match_strdup(&args[0]); 1234 if (!name) 1235 return -ENOMEM; 1236 if (!strcmp(name, "block")) { 1237 F2FS_OPTION(sbi).discard_unit = 1238 DISCARD_UNIT_BLOCK; 1239 } else if (!strcmp(name, "segment")) { 1240 F2FS_OPTION(sbi).discard_unit = 1241 DISCARD_UNIT_SEGMENT; 1242 } else if (!strcmp(name, "section")) { 1243 F2FS_OPTION(sbi).discard_unit = 1244 DISCARD_UNIT_SECTION; 1245 } else { 1246 kfree(name); 1247 return -EINVAL; 1248 } 1249 kfree(name); 1250 break; 1251 case Opt_memory_mode: 1252 name = match_strdup(&args[0]); 1253 if (!name) 1254 return -ENOMEM; 1255 if (!strcmp(name, "normal")) { 1256 F2FS_OPTION(sbi).memory_mode = 1257 MEMORY_MODE_NORMAL; 1258 } else if (!strcmp(name, "low")) { 1259 F2FS_OPTION(sbi).memory_mode = 1260 MEMORY_MODE_LOW; 1261 } else { 1262 kfree(name); 1263 return -EINVAL; 1264 } 1265 kfree(name); 1266 break; 1267 case Opt_age_extent_cache: 1268 set_opt(sbi, AGE_EXTENT_CACHE); 1269 break; 1270 case Opt_errors: 1271 name = match_strdup(&args[0]); 1272 if (!name) 1273 return -ENOMEM; 1274 if (!strcmp(name, "remount-ro")) { 1275 F2FS_OPTION(sbi).errors = 1276 MOUNT_ERRORS_READONLY; 1277 } else if (!strcmp(name, "continue")) { 1278 F2FS_OPTION(sbi).errors = 1279 MOUNT_ERRORS_CONTINUE; 1280 } else if (!strcmp(name, "panic")) { 1281 F2FS_OPTION(sbi).errors = 1282 MOUNT_ERRORS_PANIC; 1283 } else { 1284 kfree(name); 1285 return -EINVAL; 1286 } 1287 kfree(name); 1288 break; 1289 default: 1290 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 1291 p); 1292 return -EINVAL; 1293 } 1294 } 1295 default_check: 1296 #ifdef CONFIG_QUOTA 1297 if (f2fs_check_quota_options(sbi)) 1298 return -EINVAL; 1299 #else 1300 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 1301 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1302 return -EINVAL; 1303 } 1304 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 1305 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1306 return -EINVAL; 1307 } 1308 #endif 1309 #if !IS_ENABLED(CONFIG_UNICODE) 1310 if (f2fs_sb_has_casefold(sbi)) { 1311 f2fs_err(sbi, 1312 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 1313 return -EINVAL; 1314 } 1315 #endif 1316 /* 1317 * The BLKZONED feature indicates that the drive was formatted with 1318 * zone alignment optimization. This is optional for host-aware 1319 * devices, but mandatory for host-managed zoned block devices. 1320 */ 1321 if (f2fs_sb_has_blkzoned(sbi)) { 1322 #ifdef CONFIG_BLK_DEV_ZONED 1323 if (F2FS_OPTION(sbi).discard_unit != 1324 DISCARD_UNIT_SECTION) { 1325 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default"); 1326 F2FS_OPTION(sbi).discard_unit = 1327 DISCARD_UNIT_SECTION; 1328 } 1329 1330 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) { 1331 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature"); 1332 return -EINVAL; 1333 } 1334 #else 1335 f2fs_err(sbi, "Zoned block device support is not enabled"); 1336 return -EINVAL; 1337 #endif 1338 } 1339 1340 #ifdef CONFIG_F2FS_FS_COMPRESSION 1341 if (f2fs_test_compress_extension(sbi)) { 1342 f2fs_err(sbi, "invalid compress or nocompress extension"); 1343 return -EINVAL; 1344 } 1345 #endif 1346 1347 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { 1348 f2fs_err(sbi, "Should set mode=lfs with %luKB-sized IO", 1349 F2FS_IO_SIZE_KB(sbi)); 1350 return -EINVAL; 1351 } 1352 1353 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 1354 int min_size, max_size; 1355 1356 if (!f2fs_sb_has_extra_attr(sbi) || 1357 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 1358 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 1359 return -EINVAL; 1360 } 1361 if (!test_opt(sbi, INLINE_XATTR)) { 1362 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1363 return -EINVAL; 1364 } 1365 1366 min_size = MIN_INLINE_XATTR_SIZE; 1367 max_size = MAX_INLINE_XATTR_SIZE; 1368 1369 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 1370 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 1371 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 1372 min_size, max_size); 1373 return -EINVAL; 1374 } 1375 } 1376 1377 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { 1378 f2fs_err(sbi, "LFS is not compatible with checkpoint=disable"); 1379 return -EINVAL; 1380 } 1381 1382 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) { 1383 f2fs_err(sbi, "LFS is not compatible with ATGC"); 1384 return -EINVAL; 1385 } 1386 1387 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) { 1388 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode"); 1389 return -EINVAL; 1390 } 1391 1392 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) { 1393 f2fs_err(sbi, "Allow to mount readonly mode only"); 1394 return -EROFS; 1395 } 1396 return 0; 1397 } 1398 1399 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1400 { 1401 struct f2fs_inode_info *fi; 1402 1403 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) 1404 return NULL; 1405 1406 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO); 1407 if (!fi) 1408 return NULL; 1409 1410 init_once((void *) fi); 1411 1412 /* Initialize f2fs-specific inode info */ 1413 atomic_set(&fi->dirty_pages, 0); 1414 atomic_set(&fi->i_compr_blocks, 0); 1415 init_f2fs_rwsem(&fi->i_sem); 1416 spin_lock_init(&fi->i_size_lock); 1417 INIT_LIST_HEAD(&fi->dirty_list); 1418 INIT_LIST_HEAD(&fi->gdirty_list); 1419 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]); 1420 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]); 1421 init_f2fs_rwsem(&fi->i_xattr_sem); 1422 1423 /* Will be used by directory only */ 1424 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1425 1426 return &fi->vfs_inode; 1427 } 1428 1429 static int f2fs_drop_inode(struct inode *inode) 1430 { 1431 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1432 int ret; 1433 1434 /* 1435 * during filesystem shutdown, if checkpoint is disabled, 1436 * drop useless meta/node dirty pages. 1437 */ 1438 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1439 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1440 inode->i_ino == F2FS_META_INO(sbi)) { 1441 trace_f2fs_drop_inode(inode, 1); 1442 return 1; 1443 } 1444 } 1445 1446 /* 1447 * This is to avoid a deadlock condition like below. 1448 * writeback_single_inode(inode) 1449 * - f2fs_write_data_page 1450 * - f2fs_gc -> iput -> evict 1451 * - inode_wait_for_writeback(inode) 1452 */ 1453 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1454 if (!inode->i_nlink && !is_bad_inode(inode)) { 1455 /* to avoid evict_inode call simultaneously */ 1456 atomic_inc(&inode->i_count); 1457 spin_unlock(&inode->i_lock); 1458 1459 /* should remain fi->extent_tree for writepage */ 1460 f2fs_destroy_extent_node(inode); 1461 1462 sb_start_intwrite(inode->i_sb); 1463 f2fs_i_size_write(inode, 0); 1464 1465 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1466 inode, NULL, 0, DATA); 1467 truncate_inode_pages_final(inode->i_mapping); 1468 1469 if (F2FS_HAS_BLOCKS(inode)) 1470 f2fs_truncate(inode); 1471 1472 sb_end_intwrite(inode->i_sb); 1473 1474 spin_lock(&inode->i_lock); 1475 atomic_dec(&inode->i_count); 1476 } 1477 trace_f2fs_drop_inode(inode, 0); 1478 return 0; 1479 } 1480 ret = generic_drop_inode(inode); 1481 if (!ret) 1482 ret = fscrypt_drop_inode(inode); 1483 trace_f2fs_drop_inode(inode, ret); 1484 return ret; 1485 } 1486 1487 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1488 { 1489 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1490 int ret = 0; 1491 1492 spin_lock(&sbi->inode_lock[DIRTY_META]); 1493 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1494 ret = 1; 1495 } else { 1496 set_inode_flag(inode, FI_DIRTY_INODE); 1497 stat_inc_dirty_inode(sbi, DIRTY_META); 1498 } 1499 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1500 list_add_tail(&F2FS_I(inode)->gdirty_list, 1501 &sbi->inode_list[DIRTY_META]); 1502 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1503 } 1504 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1505 return ret; 1506 } 1507 1508 void f2fs_inode_synced(struct inode *inode) 1509 { 1510 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1511 1512 spin_lock(&sbi->inode_lock[DIRTY_META]); 1513 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1514 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1515 return; 1516 } 1517 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1518 list_del_init(&F2FS_I(inode)->gdirty_list); 1519 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1520 } 1521 clear_inode_flag(inode, FI_DIRTY_INODE); 1522 clear_inode_flag(inode, FI_AUTO_RECOVER); 1523 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1524 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1525 } 1526 1527 /* 1528 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1529 * 1530 * We should call set_dirty_inode to write the dirty inode through write_inode. 1531 */ 1532 static void f2fs_dirty_inode(struct inode *inode, int flags) 1533 { 1534 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1535 1536 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1537 inode->i_ino == F2FS_META_INO(sbi)) 1538 return; 1539 1540 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1541 clear_inode_flag(inode, FI_AUTO_RECOVER); 1542 1543 f2fs_inode_dirtied(inode, false); 1544 } 1545 1546 static void f2fs_free_inode(struct inode *inode) 1547 { 1548 fscrypt_free_inode(inode); 1549 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1550 } 1551 1552 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1553 { 1554 percpu_counter_destroy(&sbi->total_valid_inode_count); 1555 percpu_counter_destroy(&sbi->rf_node_block_count); 1556 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1557 } 1558 1559 static void destroy_device_list(struct f2fs_sb_info *sbi) 1560 { 1561 int i; 1562 1563 for (i = 0; i < sbi->s_ndevs; i++) { 1564 if (i > 0) 1565 blkdev_put(FDEV(i).bdev, sbi->sb->s_type); 1566 #ifdef CONFIG_BLK_DEV_ZONED 1567 kvfree(FDEV(i).blkz_seq); 1568 #endif 1569 } 1570 kvfree(sbi->devs); 1571 } 1572 1573 static void f2fs_put_super(struct super_block *sb) 1574 { 1575 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1576 int i; 1577 int err = 0; 1578 bool done; 1579 1580 /* unregister procfs/sysfs entries in advance to avoid race case */ 1581 f2fs_unregister_sysfs(sbi); 1582 1583 f2fs_quota_off_umount(sb); 1584 1585 /* prevent remaining shrinker jobs */ 1586 mutex_lock(&sbi->umount_mutex); 1587 1588 /* 1589 * flush all issued checkpoints and stop checkpoint issue thread. 1590 * after then, all checkpoints should be done by each process context. 1591 */ 1592 f2fs_stop_ckpt_thread(sbi); 1593 1594 /* 1595 * We don't need to do checkpoint when superblock is clean. 1596 * But, the previous checkpoint was not done by umount, it needs to do 1597 * clean checkpoint again. 1598 */ 1599 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1600 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1601 struct cp_control cpc = { 1602 .reason = CP_UMOUNT, 1603 }; 1604 err = f2fs_write_checkpoint(sbi, &cpc); 1605 } 1606 1607 /* be sure to wait for any on-going discard commands */ 1608 done = f2fs_issue_discard_timeout(sbi); 1609 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) { 1610 struct cp_control cpc = { 1611 .reason = CP_UMOUNT | CP_TRIMMED, 1612 }; 1613 err = f2fs_write_checkpoint(sbi, &cpc); 1614 } 1615 1616 /* 1617 * normally superblock is clean, so we need to release this. 1618 * In addition, EIO will skip do checkpoint, we need this as well. 1619 */ 1620 f2fs_release_ino_entry(sbi, true); 1621 1622 f2fs_leave_shrinker(sbi); 1623 mutex_unlock(&sbi->umount_mutex); 1624 1625 /* our cp_error case, we can wait for any writeback page */ 1626 f2fs_flush_merged_writes(sbi); 1627 1628 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1629 1630 if (err) { 1631 truncate_inode_pages_final(NODE_MAPPING(sbi)); 1632 truncate_inode_pages_final(META_MAPPING(sbi)); 1633 } 1634 1635 for (i = 0; i < NR_COUNT_TYPE; i++) { 1636 if (!get_pages(sbi, i)) 1637 continue; 1638 f2fs_err(sbi, "detect filesystem reference count leak during " 1639 "umount, type: %d, count: %lld", i, get_pages(sbi, i)); 1640 f2fs_bug_on(sbi, 1); 1641 } 1642 1643 f2fs_bug_on(sbi, sbi->fsync_node_num); 1644 1645 f2fs_destroy_compress_inode(sbi); 1646 1647 iput(sbi->node_inode); 1648 sbi->node_inode = NULL; 1649 1650 iput(sbi->meta_inode); 1651 sbi->meta_inode = NULL; 1652 1653 /* 1654 * iput() can update stat information, if f2fs_write_checkpoint() 1655 * above failed with error. 1656 */ 1657 f2fs_destroy_stats(sbi); 1658 1659 /* destroy f2fs internal modules */ 1660 f2fs_destroy_node_manager(sbi); 1661 f2fs_destroy_segment_manager(sbi); 1662 1663 /* flush s_error_work before sbi destroy */ 1664 flush_work(&sbi->s_error_work); 1665 1666 f2fs_destroy_post_read_wq(sbi); 1667 1668 kvfree(sbi->ckpt); 1669 1670 sb->s_fs_info = NULL; 1671 if (sbi->s_chksum_driver) 1672 crypto_free_shash(sbi->s_chksum_driver); 1673 kfree(sbi->raw_super); 1674 1675 destroy_device_list(sbi); 1676 f2fs_destroy_page_array_cache(sbi); 1677 f2fs_destroy_xattr_caches(sbi); 1678 mempool_destroy(sbi->write_io_dummy); 1679 #ifdef CONFIG_QUOTA 1680 for (i = 0; i < MAXQUOTAS; i++) 1681 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1682 #endif 1683 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1684 destroy_percpu_info(sbi); 1685 f2fs_destroy_iostat(sbi); 1686 for (i = 0; i < NR_PAGE_TYPE; i++) 1687 kvfree(sbi->write_io[i]); 1688 #if IS_ENABLED(CONFIG_UNICODE) 1689 utf8_unload(sb->s_encoding); 1690 #endif 1691 kfree(sbi); 1692 } 1693 1694 int f2fs_sync_fs(struct super_block *sb, int sync) 1695 { 1696 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1697 int err = 0; 1698 1699 if (unlikely(f2fs_cp_error(sbi))) 1700 return 0; 1701 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1702 return 0; 1703 1704 trace_f2fs_sync_fs(sb, sync); 1705 1706 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1707 return -EAGAIN; 1708 1709 if (sync) 1710 err = f2fs_issue_checkpoint(sbi); 1711 1712 return err; 1713 } 1714 1715 static int f2fs_freeze(struct super_block *sb) 1716 { 1717 if (f2fs_readonly(sb)) 1718 return 0; 1719 1720 /* IO error happened before */ 1721 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1722 return -EIO; 1723 1724 /* must be clean, since sync_filesystem() was already called */ 1725 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1726 return -EINVAL; 1727 1728 /* Let's flush checkpoints and stop the thread. */ 1729 f2fs_flush_ckpt_thread(F2FS_SB(sb)); 1730 1731 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */ 1732 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 1733 return 0; 1734 } 1735 1736 static int f2fs_unfreeze(struct super_block *sb) 1737 { 1738 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 1739 return 0; 1740 } 1741 1742 #ifdef CONFIG_QUOTA 1743 static int f2fs_statfs_project(struct super_block *sb, 1744 kprojid_t projid, struct kstatfs *buf) 1745 { 1746 struct kqid qid; 1747 struct dquot *dquot; 1748 u64 limit; 1749 u64 curblock; 1750 1751 qid = make_kqid_projid(projid); 1752 dquot = dqget(sb, qid); 1753 if (IS_ERR(dquot)) 1754 return PTR_ERR(dquot); 1755 spin_lock(&dquot->dq_dqb_lock); 1756 1757 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1758 dquot->dq_dqb.dqb_bhardlimit); 1759 if (limit) 1760 limit >>= sb->s_blocksize_bits; 1761 1762 if (limit && buf->f_blocks > limit) { 1763 curblock = (dquot->dq_dqb.dqb_curspace + 1764 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1765 buf->f_blocks = limit; 1766 buf->f_bfree = buf->f_bavail = 1767 (buf->f_blocks > curblock) ? 1768 (buf->f_blocks - curblock) : 0; 1769 } 1770 1771 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1772 dquot->dq_dqb.dqb_ihardlimit); 1773 1774 if (limit && buf->f_files > limit) { 1775 buf->f_files = limit; 1776 buf->f_ffree = 1777 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1778 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1779 } 1780 1781 spin_unlock(&dquot->dq_dqb_lock); 1782 dqput(dquot); 1783 return 0; 1784 } 1785 #endif 1786 1787 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1788 { 1789 struct super_block *sb = dentry->d_sb; 1790 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1791 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1792 block_t total_count, user_block_count, start_count; 1793 u64 avail_node_count; 1794 unsigned int total_valid_node_count; 1795 1796 total_count = le64_to_cpu(sbi->raw_super->block_count); 1797 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1798 buf->f_type = F2FS_SUPER_MAGIC; 1799 buf->f_bsize = sbi->blocksize; 1800 1801 buf->f_blocks = total_count - start_count; 1802 1803 spin_lock(&sbi->stat_lock); 1804 1805 user_block_count = sbi->user_block_count; 1806 total_valid_node_count = valid_node_count(sbi); 1807 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1808 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1809 sbi->current_reserved_blocks; 1810 1811 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1812 buf->f_bfree = 0; 1813 else 1814 buf->f_bfree -= sbi->unusable_block_count; 1815 spin_unlock(&sbi->stat_lock); 1816 1817 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1818 buf->f_bavail = buf->f_bfree - 1819 F2FS_OPTION(sbi).root_reserved_blocks; 1820 else 1821 buf->f_bavail = 0; 1822 1823 if (avail_node_count > user_block_count) { 1824 buf->f_files = user_block_count; 1825 buf->f_ffree = buf->f_bavail; 1826 } else { 1827 buf->f_files = avail_node_count; 1828 buf->f_ffree = min(avail_node_count - total_valid_node_count, 1829 buf->f_bavail); 1830 } 1831 1832 buf->f_namelen = F2FS_NAME_LEN; 1833 buf->f_fsid = u64_to_fsid(id); 1834 1835 #ifdef CONFIG_QUOTA 1836 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1837 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1838 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1839 } 1840 #endif 1841 return 0; 1842 } 1843 1844 static inline void f2fs_show_quota_options(struct seq_file *seq, 1845 struct super_block *sb) 1846 { 1847 #ifdef CONFIG_QUOTA 1848 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1849 1850 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1851 char *fmtname = ""; 1852 1853 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1854 case QFMT_VFS_OLD: 1855 fmtname = "vfsold"; 1856 break; 1857 case QFMT_VFS_V0: 1858 fmtname = "vfsv0"; 1859 break; 1860 case QFMT_VFS_V1: 1861 fmtname = "vfsv1"; 1862 break; 1863 } 1864 seq_printf(seq, ",jqfmt=%s", fmtname); 1865 } 1866 1867 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1868 seq_show_option(seq, "usrjquota", 1869 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1870 1871 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1872 seq_show_option(seq, "grpjquota", 1873 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1874 1875 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1876 seq_show_option(seq, "prjjquota", 1877 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1878 #endif 1879 } 1880 1881 #ifdef CONFIG_F2FS_FS_COMPRESSION 1882 static inline void f2fs_show_compress_options(struct seq_file *seq, 1883 struct super_block *sb) 1884 { 1885 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1886 char *algtype = ""; 1887 int i; 1888 1889 if (!f2fs_sb_has_compression(sbi)) 1890 return; 1891 1892 switch (F2FS_OPTION(sbi).compress_algorithm) { 1893 case COMPRESS_LZO: 1894 algtype = "lzo"; 1895 break; 1896 case COMPRESS_LZ4: 1897 algtype = "lz4"; 1898 break; 1899 case COMPRESS_ZSTD: 1900 algtype = "zstd"; 1901 break; 1902 case COMPRESS_LZORLE: 1903 algtype = "lzo-rle"; 1904 break; 1905 } 1906 seq_printf(seq, ",compress_algorithm=%s", algtype); 1907 1908 if (F2FS_OPTION(sbi).compress_level) 1909 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); 1910 1911 seq_printf(seq, ",compress_log_size=%u", 1912 F2FS_OPTION(sbi).compress_log_size); 1913 1914 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1915 seq_printf(seq, ",compress_extension=%s", 1916 F2FS_OPTION(sbi).extensions[i]); 1917 } 1918 1919 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) { 1920 seq_printf(seq, ",nocompress_extension=%s", 1921 F2FS_OPTION(sbi).noextensions[i]); 1922 } 1923 1924 if (F2FS_OPTION(sbi).compress_chksum) 1925 seq_puts(seq, ",compress_chksum"); 1926 1927 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 1928 seq_printf(seq, ",compress_mode=%s", "fs"); 1929 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 1930 seq_printf(seq, ",compress_mode=%s", "user"); 1931 1932 if (test_opt(sbi, COMPRESS_CACHE)) 1933 seq_puts(seq, ",compress_cache"); 1934 } 1935 #endif 1936 1937 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1938 { 1939 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1940 1941 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1942 seq_printf(seq, ",background_gc=%s", "sync"); 1943 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1944 seq_printf(seq, ",background_gc=%s", "on"); 1945 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1946 seq_printf(seq, ",background_gc=%s", "off"); 1947 1948 if (test_opt(sbi, GC_MERGE)) 1949 seq_puts(seq, ",gc_merge"); 1950 else 1951 seq_puts(seq, ",nogc_merge"); 1952 1953 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1954 seq_puts(seq, ",disable_roll_forward"); 1955 if (test_opt(sbi, NORECOVERY)) 1956 seq_puts(seq, ",norecovery"); 1957 if (test_opt(sbi, DISCARD)) { 1958 seq_puts(seq, ",discard"); 1959 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK) 1960 seq_printf(seq, ",discard_unit=%s", "block"); 1961 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT) 1962 seq_printf(seq, ",discard_unit=%s", "segment"); 1963 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION) 1964 seq_printf(seq, ",discard_unit=%s", "section"); 1965 } else { 1966 seq_puts(seq, ",nodiscard"); 1967 } 1968 if (test_opt(sbi, NOHEAP)) 1969 seq_puts(seq, ",no_heap"); 1970 else 1971 seq_puts(seq, ",heap"); 1972 #ifdef CONFIG_F2FS_FS_XATTR 1973 if (test_opt(sbi, XATTR_USER)) 1974 seq_puts(seq, ",user_xattr"); 1975 else 1976 seq_puts(seq, ",nouser_xattr"); 1977 if (test_opt(sbi, INLINE_XATTR)) 1978 seq_puts(seq, ",inline_xattr"); 1979 else 1980 seq_puts(seq, ",noinline_xattr"); 1981 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1982 seq_printf(seq, ",inline_xattr_size=%u", 1983 F2FS_OPTION(sbi).inline_xattr_size); 1984 #endif 1985 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1986 if (test_opt(sbi, POSIX_ACL)) 1987 seq_puts(seq, ",acl"); 1988 else 1989 seq_puts(seq, ",noacl"); 1990 #endif 1991 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1992 seq_puts(seq, ",disable_ext_identify"); 1993 if (test_opt(sbi, INLINE_DATA)) 1994 seq_puts(seq, ",inline_data"); 1995 else 1996 seq_puts(seq, ",noinline_data"); 1997 if (test_opt(sbi, INLINE_DENTRY)) 1998 seq_puts(seq, ",inline_dentry"); 1999 else 2000 seq_puts(seq, ",noinline_dentry"); 2001 if (test_opt(sbi, FLUSH_MERGE)) 2002 seq_puts(seq, ",flush_merge"); 2003 else 2004 seq_puts(seq, ",noflush_merge"); 2005 if (test_opt(sbi, NOBARRIER)) 2006 seq_puts(seq, ",nobarrier"); 2007 else 2008 seq_puts(seq, ",barrier"); 2009 if (test_opt(sbi, FASTBOOT)) 2010 seq_puts(seq, ",fastboot"); 2011 if (test_opt(sbi, READ_EXTENT_CACHE)) 2012 seq_puts(seq, ",extent_cache"); 2013 else 2014 seq_puts(seq, ",noextent_cache"); 2015 if (test_opt(sbi, AGE_EXTENT_CACHE)) 2016 seq_puts(seq, ",age_extent_cache"); 2017 if (test_opt(sbi, DATA_FLUSH)) 2018 seq_puts(seq, ",data_flush"); 2019 2020 seq_puts(seq, ",mode="); 2021 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 2022 seq_puts(seq, "adaptive"); 2023 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 2024 seq_puts(seq, "lfs"); 2025 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG) 2026 seq_puts(seq, "fragment:segment"); 2027 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK) 2028 seq_puts(seq, "fragment:block"); 2029 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 2030 if (test_opt(sbi, RESERVE_ROOT)) 2031 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 2032 F2FS_OPTION(sbi).root_reserved_blocks, 2033 from_kuid_munged(&init_user_ns, 2034 F2FS_OPTION(sbi).s_resuid), 2035 from_kgid_munged(&init_user_ns, 2036 F2FS_OPTION(sbi).s_resgid)); 2037 if (F2FS_IO_SIZE_BITS(sbi)) 2038 seq_printf(seq, ",io_bits=%u", 2039 F2FS_OPTION(sbi).write_io_size_bits); 2040 #ifdef CONFIG_F2FS_FAULT_INJECTION 2041 if (test_opt(sbi, FAULT_INJECTION)) { 2042 seq_printf(seq, ",fault_injection=%u", 2043 F2FS_OPTION(sbi).fault_info.inject_rate); 2044 seq_printf(seq, ",fault_type=%u", 2045 F2FS_OPTION(sbi).fault_info.inject_type); 2046 } 2047 #endif 2048 #ifdef CONFIG_QUOTA 2049 if (test_opt(sbi, QUOTA)) 2050 seq_puts(seq, ",quota"); 2051 if (test_opt(sbi, USRQUOTA)) 2052 seq_puts(seq, ",usrquota"); 2053 if (test_opt(sbi, GRPQUOTA)) 2054 seq_puts(seq, ",grpquota"); 2055 if (test_opt(sbi, PRJQUOTA)) 2056 seq_puts(seq, ",prjquota"); 2057 #endif 2058 f2fs_show_quota_options(seq, sbi->sb); 2059 2060 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 2061 2062 if (sbi->sb->s_flags & SB_INLINECRYPT) 2063 seq_puts(seq, ",inlinecrypt"); 2064 2065 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 2066 seq_printf(seq, ",alloc_mode=%s", "default"); 2067 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 2068 seq_printf(seq, ",alloc_mode=%s", "reuse"); 2069 2070 if (test_opt(sbi, DISABLE_CHECKPOINT)) 2071 seq_printf(seq, ",checkpoint=disable:%u", 2072 F2FS_OPTION(sbi).unusable_cap); 2073 if (test_opt(sbi, MERGE_CHECKPOINT)) 2074 seq_puts(seq, ",checkpoint_merge"); 2075 else 2076 seq_puts(seq, ",nocheckpoint_merge"); 2077 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 2078 seq_printf(seq, ",fsync_mode=%s", "posix"); 2079 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 2080 seq_printf(seq, ",fsync_mode=%s", "strict"); 2081 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 2082 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 2083 2084 #ifdef CONFIG_F2FS_FS_COMPRESSION 2085 f2fs_show_compress_options(seq, sbi->sb); 2086 #endif 2087 2088 if (test_opt(sbi, ATGC)) 2089 seq_puts(seq, ",atgc"); 2090 2091 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL) 2092 seq_printf(seq, ",memory=%s", "normal"); 2093 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW) 2094 seq_printf(seq, ",memory=%s", "low"); 2095 2096 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY) 2097 seq_printf(seq, ",errors=%s", "remount-ro"); 2098 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE) 2099 seq_printf(seq, ",errors=%s", "continue"); 2100 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC) 2101 seq_printf(seq, ",errors=%s", "panic"); 2102 2103 return 0; 2104 } 2105 2106 static void default_options(struct f2fs_sb_info *sbi, bool remount) 2107 { 2108 /* init some FS parameters */ 2109 if (!remount) { 2110 set_opt(sbi, READ_EXTENT_CACHE); 2111 clear_opt(sbi, DISABLE_CHECKPOINT); 2112 2113 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) 2114 set_opt(sbi, DISCARD); 2115 2116 if (f2fs_sb_has_blkzoned(sbi)) 2117 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION; 2118 else 2119 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK; 2120 } 2121 2122 if (f2fs_sb_has_readonly(sbi)) 2123 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE; 2124 else 2125 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 2126 2127 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 2128 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <= 2129 SMALL_VOLUME_SEGMENTS) 2130 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 2131 else 2132 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 2133 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 2134 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 2135 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 2136 if (f2fs_sb_has_compression(sbi)) { 2137 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 2138 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 2139 F2FS_OPTION(sbi).compress_ext_cnt = 0; 2140 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 2141 } 2142 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 2143 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL; 2144 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE; 2145 2146 sbi->sb->s_flags &= ~SB_INLINECRYPT; 2147 2148 set_opt(sbi, INLINE_XATTR); 2149 set_opt(sbi, INLINE_DATA); 2150 set_opt(sbi, INLINE_DENTRY); 2151 set_opt(sbi, NOHEAP); 2152 set_opt(sbi, MERGE_CHECKPOINT); 2153 F2FS_OPTION(sbi).unusable_cap = 0; 2154 sbi->sb->s_flags |= SB_LAZYTIME; 2155 if (!f2fs_is_readonly(sbi)) 2156 set_opt(sbi, FLUSH_MERGE); 2157 if (f2fs_sb_has_blkzoned(sbi)) 2158 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 2159 else 2160 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 2161 2162 #ifdef CONFIG_F2FS_FS_XATTR 2163 set_opt(sbi, XATTR_USER); 2164 #endif 2165 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2166 set_opt(sbi, POSIX_ACL); 2167 #endif 2168 2169 f2fs_build_fault_attr(sbi, 0, 0); 2170 } 2171 2172 #ifdef CONFIG_QUOTA 2173 static int f2fs_enable_quotas(struct super_block *sb); 2174 #endif 2175 2176 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 2177 { 2178 unsigned int s_flags = sbi->sb->s_flags; 2179 struct cp_control cpc; 2180 unsigned int gc_mode = sbi->gc_mode; 2181 int err = 0; 2182 int ret; 2183 block_t unusable; 2184 2185 if (s_flags & SB_RDONLY) { 2186 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 2187 return -EINVAL; 2188 } 2189 sbi->sb->s_flags |= SB_ACTIVE; 2190 2191 /* check if we need more GC first */ 2192 unusable = f2fs_get_unusable_blocks(sbi); 2193 if (!f2fs_disable_cp_again(sbi, unusable)) 2194 goto skip_gc; 2195 2196 f2fs_update_time(sbi, DISABLE_TIME); 2197 2198 sbi->gc_mode = GC_URGENT_HIGH; 2199 2200 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 2201 struct f2fs_gc_control gc_control = { 2202 .victim_segno = NULL_SEGNO, 2203 .init_gc_type = FG_GC, 2204 .should_migrate_blocks = false, 2205 .err_gc_skipped = true, 2206 .nr_free_secs = 1 }; 2207 2208 f2fs_down_write(&sbi->gc_lock); 2209 stat_inc_gc_call_count(sbi, FOREGROUND); 2210 err = f2fs_gc(sbi, &gc_control); 2211 if (err == -ENODATA) { 2212 err = 0; 2213 break; 2214 } 2215 if (err && err != -EAGAIN) 2216 break; 2217 } 2218 2219 ret = sync_filesystem(sbi->sb); 2220 if (ret || err) { 2221 err = ret ? ret : err; 2222 goto restore_flag; 2223 } 2224 2225 unusable = f2fs_get_unusable_blocks(sbi); 2226 if (f2fs_disable_cp_again(sbi, unusable)) { 2227 err = -EAGAIN; 2228 goto restore_flag; 2229 } 2230 2231 skip_gc: 2232 f2fs_down_write(&sbi->gc_lock); 2233 cpc.reason = CP_PAUSE; 2234 set_sbi_flag(sbi, SBI_CP_DISABLED); 2235 err = f2fs_write_checkpoint(sbi, &cpc); 2236 if (err) 2237 goto out_unlock; 2238 2239 spin_lock(&sbi->stat_lock); 2240 sbi->unusable_block_count = unusable; 2241 spin_unlock(&sbi->stat_lock); 2242 2243 out_unlock: 2244 f2fs_up_write(&sbi->gc_lock); 2245 restore_flag: 2246 sbi->gc_mode = gc_mode; 2247 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2248 return err; 2249 } 2250 2251 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 2252 { 2253 int retry = DEFAULT_RETRY_IO_COUNT; 2254 2255 /* we should flush all the data to keep data consistency */ 2256 do { 2257 sync_inodes_sb(sbi->sb); 2258 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2259 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--); 2260 2261 if (unlikely(retry < 0)) 2262 f2fs_warn(sbi, "checkpoint=enable has some unwritten data."); 2263 2264 f2fs_down_write(&sbi->gc_lock); 2265 f2fs_dirty_to_prefree(sbi); 2266 2267 clear_sbi_flag(sbi, SBI_CP_DISABLED); 2268 set_sbi_flag(sbi, SBI_IS_DIRTY); 2269 f2fs_up_write(&sbi->gc_lock); 2270 2271 f2fs_sync_fs(sbi->sb, 1); 2272 2273 /* Let's ensure there's no pending checkpoint anymore */ 2274 f2fs_flush_ckpt_thread(sbi); 2275 } 2276 2277 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 2278 { 2279 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2280 struct f2fs_mount_info org_mount_opt; 2281 unsigned long old_sb_flags; 2282 int err; 2283 bool need_restart_gc = false, need_stop_gc = false; 2284 bool need_restart_ckpt = false, need_stop_ckpt = false; 2285 bool need_restart_flush = false, need_stop_flush = false; 2286 bool need_restart_discard = false, need_stop_discard = false; 2287 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE); 2288 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE); 2289 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT); 2290 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 2291 bool no_atgc = !test_opt(sbi, ATGC); 2292 bool no_discard = !test_opt(sbi, DISCARD); 2293 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE); 2294 bool block_unit_discard = f2fs_block_unit_discard(sbi); 2295 #ifdef CONFIG_QUOTA 2296 int i, j; 2297 #endif 2298 2299 /* 2300 * Save the old mount options in case we 2301 * need to restore them. 2302 */ 2303 org_mount_opt = sbi->mount_opt; 2304 old_sb_flags = sb->s_flags; 2305 2306 #ifdef CONFIG_QUOTA 2307 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 2308 for (i = 0; i < MAXQUOTAS; i++) { 2309 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2310 org_mount_opt.s_qf_names[i] = 2311 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 2312 GFP_KERNEL); 2313 if (!org_mount_opt.s_qf_names[i]) { 2314 for (j = 0; j < i; j++) 2315 kfree(org_mount_opt.s_qf_names[j]); 2316 return -ENOMEM; 2317 } 2318 } else { 2319 org_mount_opt.s_qf_names[i] = NULL; 2320 } 2321 } 2322 #endif 2323 2324 /* recover superblocks we couldn't write due to previous RO mount */ 2325 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 2326 err = f2fs_commit_super(sbi, false); 2327 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 2328 err); 2329 if (!err) 2330 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2331 } 2332 2333 default_options(sbi, true); 2334 2335 /* parse mount options */ 2336 err = parse_options(sb, data, true); 2337 if (err) 2338 goto restore_opts; 2339 2340 /* flush outstanding errors before changing fs state */ 2341 flush_work(&sbi->s_error_work); 2342 2343 /* 2344 * Previous and new state of filesystem is RO, 2345 * so skip checking GC and FLUSH_MERGE conditions. 2346 */ 2347 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 2348 goto skip; 2349 2350 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) { 2351 err = -EROFS; 2352 goto restore_opts; 2353 } 2354 2355 #ifdef CONFIG_QUOTA 2356 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 2357 err = dquot_suspend(sb, -1); 2358 if (err < 0) 2359 goto restore_opts; 2360 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 2361 /* dquot_resume needs RW */ 2362 sb->s_flags &= ~SB_RDONLY; 2363 if (sb_any_quota_suspended(sb)) { 2364 dquot_resume(sb, -1); 2365 } else if (f2fs_sb_has_quota_ino(sbi)) { 2366 err = f2fs_enable_quotas(sb); 2367 if (err) 2368 goto restore_opts; 2369 } 2370 } 2371 #endif 2372 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) { 2373 err = -EINVAL; 2374 f2fs_warn(sbi, "LFS is not compatible with IPU"); 2375 goto restore_opts; 2376 } 2377 2378 /* disallow enable atgc dynamically */ 2379 if (no_atgc == !!test_opt(sbi, ATGC)) { 2380 err = -EINVAL; 2381 f2fs_warn(sbi, "switch atgc option is not allowed"); 2382 goto restore_opts; 2383 } 2384 2385 /* disallow enable/disable extent_cache dynamically */ 2386 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) { 2387 err = -EINVAL; 2388 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 2389 goto restore_opts; 2390 } 2391 /* disallow enable/disable age extent_cache dynamically */ 2392 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) { 2393 err = -EINVAL; 2394 f2fs_warn(sbi, "switch age_extent_cache option is not allowed"); 2395 goto restore_opts; 2396 } 2397 2398 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 2399 err = -EINVAL; 2400 f2fs_warn(sbi, "switch io_bits option is not allowed"); 2401 goto restore_opts; 2402 } 2403 2404 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) { 2405 err = -EINVAL; 2406 f2fs_warn(sbi, "switch compress_cache option is not allowed"); 2407 goto restore_opts; 2408 } 2409 2410 if (block_unit_discard != f2fs_block_unit_discard(sbi)) { 2411 err = -EINVAL; 2412 f2fs_warn(sbi, "switch discard_unit option is not allowed"); 2413 goto restore_opts; 2414 } 2415 2416 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 2417 err = -EINVAL; 2418 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 2419 goto restore_opts; 2420 } 2421 2422 /* 2423 * We stop the GC thread if FS is mounted as RO 2424 * or if background_gc = off is passed in mount 2425 * option. Also sync the filesystem. 2426 */ 2427 if ((*flags & SB_RDONLY) || 2428 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF && 2429 !test_opt(sbi, GC_MERGE))) { 2430 if (sbi->gc_thread) { 2431 f2fs_stop_gc_thread(sbi); 2432 need_restart_gc = true; 2433 } 2434 } else if (!sbi->gc_thread) { 2435 err = f2fs_start_gc_thread(sbi); 2436 if (err) 2437 goto restore_opts; 2438 need_stop_gc = true; 2439 } 2440 2441 if (*flags & SB_RDONLY) { 2442 sync_inodes_sb(sb); 2443 2444 set_sbi_flag(sbi, SBI_IS_DIRTY); 2445 set_sbi_flag(sbi, SBI_IS_CLOSE); 2446 f2fs_sync_fs(sb, 1); 2447 clear_sbi_flag(sbi, SBI_IS_CLOSE); 2448 } 2449 2450 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) || 2451 !test_opt(sbi, MERGE_CHECKPOINT)) { 2452 f2fs_stop_ckpt_thread(sbi); 2453 need_restart_ckpt = true; 2454 } else { 2455 /* Flush if the prevous checkpoint, if exists. */ 2456 f2fs_flush_ckpt_thread(sbi); 2457 2458 err = f2fs_start_ckpt_thread(sbi); 2459 if (err) { 2460 f2fs_err(sbi, 2461 "Failed to start F2FS issue_checkpoint_thread (%d)", 2462 err); 2463 goto restore_gc; 2464 } 2465 need_stop_ckpt = true; 2466 } 2467 2468 /* 2469 * We stop issue flush thread if FS is mounted as RO 2470 * or if flush_merge is not passed in mount option. 2471 */ 2472 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 2473 clear_opt(sbi, FLUSH_MERGE); 2474 f2fs_destroy_flush_cmd_control(sbi, false); 2475 need_restart_flush = true; 2476 } else { 2477 err = f2fs_create_flush_cmd_control(sbi); 2478 if (err) 2479 goto restore_ckpt; 2480 need_stop_flush = true; 2481 } 2482 2483 if (no_discard == !!test_opt(sbi, DISCARD)) { 2484 if (test_opt(sbi, DISCARD)) { 2485 err = f2fs_start_discard_thread(sbi); 2486 if (err) 2487 goto restore_flush; 2488 need_stop_discard = true; 2489 } else { 2490 f2fs_stop_discard_thread(sbi); 2491 f2fs_issue_discard_timeout(sbi); 2492 need_restart_discard = true; 2493 } 2494 } 2495 2496 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) { 2497 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 2498 err = f2fs_disable_checkpoint(sbi); 2499 if (err) 2500 goto restore_discard; 2501 } else { 2502 f2fs_enable_checkpoint(sbi); 2503 } 2504 } 2505 2506 skip: 2507 #ifdef CONFIG_QUOTA 2508 /* Release old quota file names */ 2509 for (i = 0; i < MAXQUOTAS; i++) 2510 kfree(org_mount_opt.s_qf_names[i]); 2511 #endif 2512 /* Update the POSIXACL Flag */ 2513 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2514 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2515 2516 limit_reserve_root(sbi); 2517 adjust_unusable_cap_perc(sbi); 2518 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 2519 return 0; 2520 restore_discard: 2521 if (need_restart_discard) { 2522 if (f2fs_start_discard_thread(sbi)) 2523 f2fs_warn(sbi, "discard has been stopped"); 2524 } else if (need_stop_discard) { 2525 f2fs_stop_discard_thread(sbi); 2526 } 2527 restore_flush: 2528 if (need_restart_flush) { 2529 if (f2fs_create_flush_cmd_control(sbi)) 2530 f2fs_warn(sbi, "background flush thread has stopped"); 2531 } else if (need_stop_flush) { 2532 clear_opt(sbi, FLUSH_MERGE); 2533 f2fs_destroy_flush_cmd_control(sbi, false); 2534 } 2535 restore_ckpt: 2536 if (need_restart_ckpt) { 2537 if (f2fs_start_ckpt_thread(sbi)) 2538 f2fs_warn(sbi, "background ckpt thread has stopped"); 2539 } else if (need_stop_ckpt) { 2540 f2fs_stop_ckpt_thread(sbi); 2541 } 2542 restore_gc: 2543 if (need_restart_gc) { 2544 if (f2fs_start_gc_thread(sbi)) 2545 f2fs_warn(sbi, "background gc thread has stopped"); 2546 } else if (need_stop_gc) { 2547 f2fs_stop_gc_thread(sbi); 2548 } 2549 restore_opts: 2550 #ifdef CONFIG_QUOTA 2551 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 2552 for (i = 0; i < MAXQUOTAS; i++) { 2553 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2554 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 2555 } 2556 #endif 2557 sbi->mount_opt = org_mount_opt; 2558 sb->s_flags = old_sb_flags; 2559 return err; 2560 } 2561 2562 #ifdef CONFIG_QUOTA 2563 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi) 2564 { 2565 /* need to recovery orphan */ 2566 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG)) 2567 return true; 2568 /* need to recovery data */ 2569 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 2570 return false; 2571 if (test_opt(sbi, NORECOVERY)) 2572 return false; 2573 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG); 2574 } 2575 2576 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi) 2577 { 2578 bool readonly = f2fs_readonly(sbi->sb); 2579 2580 if (!f2fs_need_recovery(sbi)) 2581 return false; 2582 2583 /* it doesn't need to check f2fs_sb_has_readonly() */ 2584 if (f2fs_hw_is_readonly(sbi)) 2585 return false; 2586 2587 if (readonly) { 2588 sbi->sb->s_flags &= ~SB_RDONLY; 2589 set_sbi_flag(sbi, SBI_IS_WRITABLE); 2590 } 2591 2592 /* 2593 * Turn on quotas which were not enabled for read-only mounts if 2594 * filesystem has quota feature, so that they are updated correctly. 2595 */ 2596 return f2fs_enable_quota_files(sbi, readonly); 2597 } 2598 2599 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi, 2600 bool quota_enabled) 2601 { 2602 if (quota_enabled) 2603 f2fs_quota_off_umount(sbi->sb); 2604 2605 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) { 2606 clear_sbi_flag(sbi, SBI_IS_WRITABLE); 2607 sbi->sb->s_flags |= SB_RDONLY; 2608 } 2609 } 2610 2611 /* Read data from quotafile */ 2612 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 2613 size_t len, loff_t off) 2614 { 2615 struct inode *inode = sb_dqopt(sb)->files[type]; 2616 struct address_space *mapping = inode->i_mapping; 2617 block_t blkidx = F2FS_BYTES_TO_BLK(off); 2618 int offset = off & (sb->s_blocksize - 1); 2619 int tocopy; 2620 size_t toread; 2621 loff_t i_size = i_size_read(inode); 2622 struct page *page; 2623 2624 if (off > i_size) 2625 return 0; 2626 2627 if (off + len > i_size) 2628 len = i_size - off; 2629 toread = len; 2630 while (toread > 0) { 2631 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 2632 repeat: 2633 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 2634 if (IS_ERR(page)) { 2635 if (PTR_ERR(page) == -ENOMEM) { 2636 memalloc_retry_wait(GFP_NOFS); 2637 goto repeat; 2638 } 2639 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2640 return PTR_ERR(page); 2641 } 2642 2643 lock_page(page); 2644 2645 if (unlikely(page->mapping != mapping)) { 2646 f2fs_put_page(page, 1); 2647 goto repeat; 2648 } 2649 if (unlikely(!PageUptodate(page))) { 2650 f2fs_put_page(page, 1); 2651 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2652 return -EIO; 2653 } 2654 2655 memcpy_from_page(data, page, offset, tocopy); 2656 f2fs_put_page(page, 1); 2657 2658 offset = 0; 2659 toread -= tocopy; 2660 data += tocopy; 2661 blkidx++; 2662 } 2663 return len; 2664 } 2665 2666 /* Write to quotafile */ 2667 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2668 const char *data, size_t len, loff_t off) 2669 { 2670 struct inode *inode = sb_dqopt(sb)->files[type]; 2671 struct address_space *mapping = inode->i_mapping; 2672 const struct address_space_operations *a_ops = mapping->a_ops; 2673 int offset = off & (sb->s_blocksize - 1); 2674 size_t towrite = len; 2675 struct page *page; 2676 void *fsdata = NULL; 2677 int err = 0; 2678 int tocopy; 2679 2680 while (towrite > 0) { 2681 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2682 towrite); 2683 retry: 2684 err = a_ops->write_begin(NULL, mapping, off, tocopy, 2685 &page, &fsdata); 2686 if (unlikely(err)) { 2687 if (err == -ENOMEM) { 2688 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2689 goto retry; 2690 } 2691 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2692 break; 2693 } 2694 2695 memcpy_to_page(page, offset, data, tocopy); 2696 2697 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2698 page, fsdata); 2699 offset = 0; 2700 towrite -= tocopy; 2701 off += tocopy; 2702 data += tocopy; 2703 cond_resched(); 2704 } 2705 2706 if (len == towrite) 2707 return err; 2708 inode->i_mtime = inode->i_ctime = current_time(inode); 2709 f2fs_mark_inode_dirty_sync(inode, false); 2710 return len - towrite; 2711 } 2712 2713 int f2fs_dquot_initialize(struct inode *inode) 2714 { 2715 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) 2716 return -ESRCH; 2717 2718 return dquot_initialize(inode); 2719 } 2720 2721 static struct dquot **f2fs_get_dquots(struct inode *inode) 2722 { 2723 return F2FS_I(inode)->i_dquot; 2724 } 2725 2726 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2727 { 2728 return &F2FS_I(inode)->i_reserved_quota; 2729 } 2730 2731 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2732 { 2733 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2734 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2735 return 0; 2736 } 2737 2738 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2739 F2FS_OPTION(sbi).s_jquota_fmt, type); 2740 } 2741 2742 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2743 { 2744 int enabled = 0; 2745 int i, err; 2746 2747 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2748 err = f2fs_enable_quotas(sbi->sb); 2749 if (err) { 2750 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2751 return 0; 2752 } 2753 return 1; 2754 } 2755 2756 for (i = 0; i < MAXQUOTAS; i++) { 2757 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2758 err = f2fs_quota_on_mount(sbi, i); 2759 if (!err) { 2760 enabled = 1; 2761 continue; 2762 } 2763 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2764 err, i); 2765 } 2766 } 2767 return enabled; 2768 } 2769 2770 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2771 unsigned int flags) 2772 { 2773 struct inode *qf_inode; 2774 unsigned long qf_inum; 2775 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL; 2776 int err; 2777 2778 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2779 2780 qf_inum = f2fs_qf_ino(sb, type); 2781 if (!qf_inum) 2782 return -EPERM; 2783 2784 qf_inode = f2fs_iget(sb, qf_inum); 2785 if (IS_ERR(qf_inode)) { 2786 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2787 return PTR_ERR(qf_inode); 2788 } 2789 2790 /* Don't account quota for quota files to avoid recursion */ 2791 inode_lock(qf_inode); 2792 qf_inode->i_flags |= S_NOQUOTA; 2793 2794 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) { 2795 F2FS_I(qf_inode)->i_flags |= qf_flag; 2796 f2fs_set_inode_flags(qf_inode); 2797 } 2798 inode_unlock(qf_inode); 2799 2800 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2801 iput(qf_inode); 2802 return err; 2803 } 2804 2805 static int f2fs_enable_quotas(struct super_block *sb) 2806 { 2807 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2808 int type, err = 0; 2809 unsigned long qf_inum; 2810 bool quota_mopt[MAXQUOTAS] = { 2811 test_opt(sbi, USRQUOTA), 2812 test_opt(sbi, GRPQUOTA), 2813 test_opt(sbi, PRJQUOTA), 2814 }; 2815 2816 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2817 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2818 return 0; 2819 } 2820 2821 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2822 2823 for (type = 0; type < MAXQUOTAS; type++) { 2824 qf_inum = f2fs_qf_ino(sb, type); 2825 if (qf_inum) { 2826 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2827 DQUOT_USAGE_ENABLED | 2828 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2829 if (err) { 2830 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2831 type, err); 2832 for (type--; type >= 0; type--) 2833 dquot_quota_off(sb, type); 2834 set_sbi_flag(F2FS_SB(sb), 2835 SBI_QUOTA_NEED_REPAIR); 2836 return err; 2837 } 2838 } 2839 } 2840 return 0; 2841 } 2842 2843 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type) 2844 { 2845 struct quota_info *dqopt = sb_dqopt(sbi->sb); 2846 struct address_space *mapping = dqopt->files[type]->i_mapping; 2847 int ret = 0; 2848 2849 ret = dquot_writeback_dquots(sbi->sb, type); 2850 if (ret) 2851 goto out; 2852 2853 ret = filemap_fdatawrite(mapping); 2854 if (ret) 2855 goto out; 2856 2857 /* if we are using journalled quota */ 2858 if (is_journalled_quota(sbi)) 2859 goto out; 2860 2861 ret = filemap_fdatawait(mapping); 2862 2863 truncate_inode_pages(&dqopt->files[type]->i_data, 0); 2864 out: 2865 if (ret) 2866 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2867 return ret; 2868 } 2869 2870 int f2fs_quota_sync(struct super_block *sb, int type) 2871 { 2872 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2873 struct quota_info *dqopt = sb_dqopt(sb); 2874 int cnt; 2875 int ret = 0; 2876 2877 /* 2878 * Now when everything is written we can discard the pagecache so 2879 * that userspace sees the changes. 2880 */ 2881 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2882 2883 if (type != -1 && cnt != type) 2884 continue; 2885 2886 if (!sb_has_quota_active(sb, cnt)) 2887 continue; 2888 2889 if (!f2fs_sb_has_quota_ino(sbi)) 2890 inode_lock(dqopt->files[cnt]); 2891 2892 /* 2893 * do_quotactl 2894 * f2fs_quota_sync 2895 * f2fs_down_read(quota_sem) 2896 * dquot_writeback_dquots() 2897 * f2fs_dquot_commit 2898 * block_operation 2899 * f2fs_down_read(quota_sem) 2900 */ 2901 f2fs_lock_op(sbi); 2902 f2fs_down_read(&sbi->quota_sem); 2903 2904 ret = f2fs_quota_sync_file(sbi, cnt); 2905 2906 f2fs_up_read(&sbi->quota_sem); 2907 f2fs_unlock_op(sbi); 2908 2909 if (!f2fs_sb_has_quota_ino(sbi)) 2910 inode_unlock(dqopt->files[cnt]); 2911 2912 if (ret) 2913 break; 2914 } 2915 return ret; 2916 } 2917 2918 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2919 const struct path *path) 2920 { 2921 struct inode *inode; 2922 int err; 2923 2924 /* if quota sysfile exists, deny enabling quota with specific file */ 2925 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2926 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2927 return -EBUSY; 2928 } 2929 2930 if (path->dentry->d_sb != sb) 2931 return -EXDEV; 2932 2933 err = f2fs_quota_sync(sb, type); 2934 if (err) 2935 return err; 2936 2937 inode = d_inode(path->dentry); 2938 2939 err = filemap_fdatawrite(inode->i_mapping); 2940 if (err) 2941 return err; 2942 2943 err = filemap_fdatawait(inode->i_mapping); 2944 if (err) 2945 return err; 2946 2947 err = dquot_quota_on(sb, type, format_id, path); 2948 if (err) 2949 return err; 2950 2951 inode_lock(inode); 2952 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL; 2953 f2fs_set_inode_flags(inode); 2954 inode_unlock(inode); 2955 f2fs_mark_inode_dirty_sync(inode, false); 2956 2957 return 0; 2958 } 2959 2960 static int __f2fs_quota_off(struct super_block *sb, int type) 2961 { 2962 struct inode *inode = sb_dqopt(sb)->files[type]; 2963 int err; 2964 2965 if (!inode || !igrab(inode)) 2966 return dquot_quota_off(sb, type); 2967 2968 err = f2fs_quota_sync(sb, type); 2969 if (err) 2970 goto out_put; 2971 2972 err = dquot_quota_off(sb, type); 2973 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2974 goto out_put; 2975 2976 inode_lock(inode); 2977 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL; 2978 f2fs_set_inode_flags(inode); 2979 inode_unlock(inode); 2980 f2fs_mark_inode_dirty_sync(inode, false); 2981 out_put: 2982 iput(inode); 2983 return err; 2984 } 2985 2986 static int f2fs_quota_off(struct super_block *sb, int type) 2987 { 2988 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2989 int err; 2990 2991 err = __f2fs_quota_off(sb, type); 2992 2993 /* 2994 * quotactl can shutdown journalled quota, result in inconsistence 2995 * between quota record and fs data by following updates, tag the 2996 * flag to let fsck be aware of it. 2997 */ 2998 if (is_journalled_quota(sbi)) 2999 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3000 return err; 3001 } 3002 3003 void f2fs_quota_off_umount(struct super_block *sb) 3004 { 3005 int type; 3006 int err; 3007 3008 for (type = 0; type < MAXQUOTAS; type++) { 3009 err = __f2fs_quota_off(sb, type); 3010 if (err) { 3011 int ret = dquot_quota_off(sb, type); 3012 3013 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 3014 type, err, ret); 3015 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 3016 } 3017 } 3018 /* 3019 * In case of checkpoint=disable, we must flush quota blocks. 3020 * This can cause NULL exception for node_inode in end_io, since 3021 * put_super already dropped it. 3022 */ 3023 sync_filesystem(sb); 3024 } 3025 3026 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 3027 { 3028 struct quota_info *dqopt = sb_dqopt(sb); 3029 int type; 3030 3031 for (type = 0; type < MAXQUOTAS; type++) { 3032 if (!dqopt->files[type]) 3033 continue; 3034 f2fs_inode_synced(dqopt->files[type]); 3035 } 3036 } 3037 3038 static int f2fs_dquot_commit(struct dquot *dquot) 3039 { 3040 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3041 int ret; 3042 3043 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 3044 ret = dquot_commit(dquot); 3045 if (ret < 0) 3046 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3047 f2fs_up_read(&sbi->quota_sem); 3048 return ret; 3049 } 3050 3051 static int f2fs_dquot_acquire(struct dquot *dquot) 3052 { 3053 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3054 int ret; 3055 3056 f2fs_down_read(&sbi->quota_sem); 3057 ret = dquot_acquire(dquot); 3058 if (ret < 0) 3059 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3060 f2fs_up_read(&sbi->quota_sem); 3061 return ret; 3062 } 3063 3064 static int f2fs_dquot_release(struct dquot *dquot) 3065 { 3066 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3067 int ret = dquot_release(dquot); 3068 3069 if (ret < 0) 3070 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3071 return ret; 3072 } 3073 3074 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 3075 { 3076 struct super_block *sb = dquot->dq_sb; 3077 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3078 int ret = dquot_mark_dquot_dirty(dquot); 3079 3080 /* if we are using journalled quota */ 3081 if (is_journalled_quota(sbi)) 3082 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 3083 3084 return ret; 3085 } 3086 3087 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 3088 { 3089 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3090 int ret = dquot_commit_info(sb, type); 3091 3092 if (ret < 0) 3093 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3094 return ret; 3095 } 3096 3097 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 3098 { 3099 *projid = F2FS_I(inode)->i_projid; 3100 return 0; 3101 } 3102 3103 static const struct dquot_operations f2fs_quota_operations = { 3104 .get_reserved_space = f2fs_get_reserved_space, 3105 .write_dquot = f2fs_dquot_commit, 3106 .acquire_dquot = f2fs_dquot_acquire, 3107 .release_dquot = f2fs_dquot_release, 3108 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 3109 .write_info = f2fs_dquot_commit_info, 3110 .alloc_dquot = dquot_alloc, 3111 .destroy_dquot = dquot_destroy, 3112 .get_projid = f2fs_get_projid, 3113 .get_next_id = dquot_get_next_id, 3114 }; 3115 3116 static const struct quotactl_ops f2fs_quotactl_ops = { 3117 .quota_on = f2fs_quota_on, 3118 .quota_off = f2fs_quota_off, 3119 .quota_sync = f2fs_quota_sync, 3120 .get_state = dquot_get_state, 3121 .set_info = dquot_set_dqinfo, 3122 .get_dqblk = dquot_get_dqblk, 3123 .set_dqblk = dquot_set_dqblk, 3124 .get_nextdqblk = dquot_get_next_dqblk, 3125 }; 3126 #else 3127 int f2fs_dquot_initialize(struct inode *inode) 3128 { 3129 return 0; 3130 } 3131 3132 int f2fs_quota_sync(struct super_block *sb, int type) 3133 { 3134 return 0; 3135 } 3136 3137 void f2fs_quota_off_umount(struct super_block *sb) 3138 { 3139 } 3140 #endif 3141 3142 static const struct super_operations f2fs_sops = { 3143 .alloc_inode = f2fs_alloc_inode, 3144 .free_inode = f2fs_free_inode, 3145 .drop_inode = f2fs_drop_inode, 3146 .write_inode = f2fs_write_inode, 3147 .dirty_inode = f2fs_dirty_inode, 3148 .show_options = f2fs_show_options, 3149 #ifdef CONFIG_QUOTA 3150 .quota_read = f2fs_quota_read, 3151 .quota_write = f2fs_quota_write, 3152 .get_dquots = f2fs_get_dquots, 3153 #endif 3154 .evict_inode = f2fs_evict_inode, 3155 .put_super = f2fs_put_super, 3156 .sync_fs = f2fs_sync_fs, 3157 .freeze_fs = f2fs_freeze, 3158 .unfreeze_fs = f2fs_unfreeze, 3159 .statfs = f2fs_statfs, 3160 .remount_fs = f2fs_remount, 3161 }; 3162 3163 #ifdef CONFIG_FS_ENCRYPTION 3164 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 3165 { 3166 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3167 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3168 ctx, len, NULL); 3169 } 3170 3171 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 3172 void *fs_data) 3173 { 3174 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3175 3176 /* 3177 * Encrypting the root directory is not allowed because fsck 3178 * expects lost+found directory to exist and remain unencrypted 3179 * if LOST_FOUND feature is enabled. 3180 * 3181 */ 3182 if (f2fs_sb_has_lost_found(sbi) && 3183 inode->i_ino == F2FS_ROOT_INO(sbi)) 3184 return -EPERM; 3185 3186 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3187 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3188 ctx, len, fs_data, XATTR_CREATE); 3189 } 3190 3191 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 3192 { 3193 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 3194 } 3195 3196 static bool f2fs_has_stable_inodes(struct super_block *sb) 3197 { 3198 return true; 3199 } 3200 3201 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 3202 int *ino_bits_ret, int *lblk_bits_ret) 3203 { 3204 *ino_bits_ret = 8 * sizeof(nid_t); 3205 *lblk_bits_ret = 8 * sizeof(block_t); 3206 } 3207 3208 static struct block_device **f2fs_get_devices(struct super_block *sb, 3209 unsigned int *num_devs) 3210 { 3211 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3212 struct block_device **devs; 3213 int i; 3214 3215 if (!f2fs_is_multi_device(sbi)) 3216 return NULL; 3217 3218 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL); 3219 if (!devs) 3220 return ERR_PTR(-ENOMEM); 3221 3222 for (i = 0; i < sbi->s_ndevs; i++) 3223 devs[i] = FDEV(i).bdev; 3224 *num_devs = sbi->s_ndevs; 3225 return devs; 3226 } 3227 3228 static const struct fscrypt_operations f2fs_cryptops = { 3229 .key_prefix = "f2fs:", 3230 .get_context = f2fs_get_context, 3231 .set_context = f2fs_set_context, 3232 .get_dummy_policy = f2fs_get_dummy_policy, 3233 .empty_dir = f2fs_empty_dir, 3234 .has_stable_inodes = f2fs_has_stable_inodes, 3235 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 3236 .get_devices = f2fs_get_devices, 3237 }; 3238 #endif 3239 3240 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 3241 u64 ino, u32 generation) 3242 { 3243 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3244 struct inode *inode; 3245 3246 if (f2fs_check_nid_range(sbi, ino)) 3247 return ERR_PTR(-ESTALE); 3248 3249 /* 3250 * f2fs_iget isn't quite right if the inode is currently unallocated! 3251 * However f2fs_iget currently does appropriate checks to handle stale 3252 * inodes so everything is OK. 3253 */ 3254 inode = f2fs_iget(sb, ino); 3255 if (IS_ERR(inode)) 3256 return ERR_CAST(inode); 3257 if (unlikely(generation && inode->i_generation != generation)) { 3258 /* we didn't find the right inode.. */ 3259 iput(inode); 3260 return ERR_PTR(-ESTALE); 3261 } 3262 return inode; 3263 } 3264 3265 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 3266 int fh_len, int fh_type) 3267 { 3268 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 3269 f2fs_nfs_get_inode); 3270 } 3271 3272 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 3273 int fh_len, int fh_type) 3274 { 3275 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 3276 f2fs_nfs_get_inode); 3277 } 3278 3279 static const struct export_operations f2fs_export_ops = { 3280 .fh_to_dentry = f2fs_fh_to_dentry, 3281 .fh_to_parent = f2fs_fh_to_parent, 3282 .get_parent = f2fs_get_parent, 3283 }; 3284 3285 loff_t max_file_blocks(struct inode *inode) 3286 { 3287 loff_t result = 0; 3288 loff_t leaf_count; 3289 3290 /* 3291 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 3292 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 3293 * space in inode.i_addr, it will be more safe to reassign 3294 * result as zero. 3295 */ 3296 3297 if (inode && f2fs_compressed_file(inode)) 3298 leaf_count = ADDRS_PER_BLOCK(inode); 3299 else 3300 leaf_count = DEF_ADDRS_PER_BLOCK; 3301 3302 /* two direct node blocks */ 3303 result += (leaf_count * 2); 3304 3305 /* two indirect node blocks */ 3306 leaf_count *= NIDS_PER_BLOCK; 3307 result += (leaf_count * 2); 3308 3309 /* one double indirect node block */ 3310 leaf_count *= NIDS_PER_BLOCK; 3311 result += leaf_count; 3312 3313 return result; 3314 } 3315 3316 static int __f2fs_commit_super(struct buffer_head *bh, 3317 struct f2fs_super_block *super) 3318 { 3319 lock_buffer(bh); 3320 if (super) 3321 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 3322 set_buffer_dirty(bh); 3323 unlock_buffer(bh); 3324 3325 /* it's rare case, we can do fua all the time */ 3326 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 3327 } 3328 3329 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 3330 struct buffer_head *bh) 3331 { 3332 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3333 (bh->b_data + F2FS_SUPER_OFFSET); 3334 struct super_block *sb = sbi->sb; 3335 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 3336 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 3337 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 3338 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 3339 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 3340 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 3341 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 3342 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 3343 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 3344 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 3345 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3346 u32 segment_count = le32_to_cpu(raw_super->segment_count); 3347 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3348 u64 main_end_blkaddr = main_blkaddr + 3349 (segment_count_main << log_blocks_per_seg); 3350 u64 seg_end_blkaddr = segment0_blkaddr + 3351 (segment_count << log_blocks_per_seg); 3352 3353 if (segment0_blkaddr != cp_blkaddr) { 3354 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 3355 segment0_blkaddr, cp_blkaddr); 3356 return true; 3357 } 3358 3359 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 3360 sit_blkaddr) { 3361 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 3362 cp_blkaddr, sit_blkaddr, 3363 segment_count_ckpt << log_blocks_per_seg); 3364 return true; 3365 } 3366 3367 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 3368 nat_blkaddr) { 3369 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 3370 sit_blkaddr, nat_blkaddr, 3371 segment_count_sit << log_blocks_per_seg); 3372 return true; 3373 } 3374 3375 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 3376 ssa_blkaddr) { 3377 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 3378 nat_blkaddr, ssa_blkaddr, 3379 segment_count_nat << log_blocks_per_seg); 3380 return true; 3381 } 3382 3383 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 3384 main_blkaddr) { 3385 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 3386 ssa_blkaddr, main_blkaddr, 3387 segment_count_ssa << log_blocks_per_seg); 3388 return true; 3389 } 3390 3391 if (main_end_blkaddr > seg_end_blkaddr) { 3392 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 3393 main_blkaddr, seg_end_blkaddr, 3394 segment_count_main << log_blocks_per_seg); 3395 return true; 3396 } else if (main_end_blkaddr < seg_end_blkaddr) { 3397 int err = 0; 3398 char *res; 3399 3400 /* fix in-memory information all the time */ 3401 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 3402 segment0_blkaddr) >> log_blocks_per_seg); 3403 3404 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) { 3405 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3406 res = "internally"; 3407 } else { 3408 err = __f2fs_commit_super(bh, NULL); 3409 res = err ? "failed" : "done"; 3410 } 3411 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 3412 res, main_blkaddr, seg_end_blkaddr, 3413 segment_count_main << log_blocks_per_seg); 3414 if (err) 3415 return true; 3416 } 3417 return false; 3418 } 3419 3420 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 3421 struct buffer_head *bh) 3422 { 3423 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 3424 block_t total_sections, blocks_per_seg; 3425 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3426 (bh->b_data + F2FS_SUPER_OFFSET); 3427 size_t crc_offset = 0; 3428 __u32 crc = 0; 3429 3430 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 3431 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 3432 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 3433 return -EINVAL; 3434 } 3435 3436 /* Check checksum_offset and crc in superblock */ 3437 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 3438 crc_offset = le32_to_cpu(raw_super->checksum_offset); 3439 if (crc_offset != 3440 offsetof(struct f2fs_super_block, crc)) { 3441 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 3442 crc_offset); 3443 return -EFSCORRUPTED; 3444 } 3445 crc = le32_to_cpu(raw_super->crc); 3446 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 3447 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 3448 return -EFSCORRUPTED; 3449 } 3450 } 3451 3452 /* Currently, support only 4KB block size */ 3453 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 3454 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 3455 le32_to_cpu(raw_super->log_blocksize), 3456 F2FS_BLKSIZE_BITS); 3457 return -EFSCORRUPTED; 3458 } 3459 3460 /* check log blocks per segment */ 3461 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 3462 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 3463 le32_to_cpu(raw_super->log_blocks_per_seg)); 3464 return -EFSCORRUPTED; 3465 } 3466 3467 /* Currently, support 512/1024/2048/4096 bytes sector size */ 3468 if (le32_to_cpu(raw_super->log_sectorsize) > 3469 F2FS_MAX_LOG_SECTOR_SIZE || 3470 le32_to_cpu(raw_super->log_sectorsize) < 3471 F2FS_MIN_LOG_SECTOR_SIZE) { 3472 f2fs_info(sbi, "Invalid log sectorsize (%u)", 3473 le32_to_cpu(raw_super->log_sectorsize)); 3474 return -EFSCORRUPTED; 3475 } 3476 if (le32_to_cpu(raw_super->log_sectors_per_block) + 3477 le32_to_cpu(raw_super->log_sectorsize) != 3478 F2FS_MAX_LOG_SECTOR_SIZE) { 3479 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 3480 le32_to_cpu(raw_super->log_sectors_per_block), 3481 le32_to_cpu(raw_super->log_sectorsize)); 3482 return -EFSCORRUPTED; 3483 } 3484 3485 segment_count = le32_to_cpu(raw_super->segment_count); 3486 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3487 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3488 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3489 total_sections = le32_to_cpu(raw_super->section_count); 3490 3491 /* blocks_per_seg should be 512, given the above check */ 3492 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg)); 3493 3494 if (segment_count > F2FS_MAX_SEGMENT || 3495 segment_count < F2FS_MIN_SEGMENTS) { 3496 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 3497 return -EFSCORRUPTED; 3498 } 3499 3500 if (total_sections > segment_count_main || total_sections < 1 || 3501 segs_per_sec > segment_count || !segs_per_sec) { 3502 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 3503 segment_count, total_sections, segs_per_sec); 3504 return -EFSCORRUPTED; 3505 } 3506 3507 if (segment_count_main != total_sections * segs_per_sec) { 3508 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 3509 segment_count_main, total_sections, segs_per_sec); 3510 return -EFSCORRUPTED; 3511 } 3512 3513 if ((segment_count / segs_per_sec) < total_sections) { 3514 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 3515 segment_count, segs_per_sec, total_sections); 3516 return -EFSCORRUPTED; 3517 } 3518 3519 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 3520 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 3521 segment_count, le64_to_cpu(raw_super->block_count)); 3522 return -EFSCORRUPTED; 3523 } 3524 3525 if (RDEV(0).path[0]) { 3526 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 3527 int i = 1; 3528 3529 while (i < MAX_DEVICES && RDEV(i).path[0]) { 3530 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 3531 i++; 3532 } 3533 if (segment_count != dev_seg_count) { 3534 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 3535 segment_count, dev_seg_count); 3536 return -EFSCORRUPTED; 3537 } 3538 } else { 3539 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 3540 !bdev_is_zoned(sbi->sb->s_bdev)) { 3541 f2fs_info(sbi, "Zoned block device path is missing"); 3542 return -EFSCORRUPTED; 3543 } 3544 } 3545 3546 if (secs_per_zone > total_sections || !secs_per_zone) { 3547 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 3548 secs_per_zone, total_sections); 3549 return -EFSCORRUPTED; 3550 } 3551 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 3552 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 3553 (le32_to_cpu(raw_super->extension_count) + 3554 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 3555 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 3556 le32_to_cpu(raw_super->extension_count), 3557 raw_super->hot_ext_count, 3558 F2FS_MAX_EXTENSION); 3559 return -EFSCORRUPTED; 3560 } 3561 3562 if (le32_to_cpu(raw_super->cp_payload) >= 3563 (blocks_per_seg - F2FS_CP_PACKS - 3564 NR_CURSEG_PERSIST_TYPE)) { 3565 f2fs_info(sbi, "Insane cp_payload (%u >= %u)", 3566 le32_to_cpu(raw_super->cp_payload), 3567 blocks_per_seg - F2FS_CP_PACKS - 3568 NR_CURSEG_PERSIST_TYPE); 3569 return -EFSCORRUPTED; 3570 } 3571 3572 /* check reserved ino info */ 3573 if (le32_to_cpu(raw_super->node_ino) != 1 || 3574 le32_to_cpu(raw_super->meta_ino) != 2 || 3575 le32_to_cpu(raw_super->root_ino) != 3) { 3576 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 3577 le32_to_cpu(raw_super->node_ino), 3578 le32_to_cpu(raw_super->meta_ino), 3579 le32_to_cpu(raw_super->root_ino)); 3580 return -EFSCORRUPTED; 3581 } 3582 3583 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 3584 if (sanity_check_area_boundary(sbi, bh)) 3585 return -EFSCORRUPTED; 3586 3587 return 0; 3588 } 3589 3590 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 3591 { 3592 unsigned int total, fsmeta; 3593 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3594 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 3595 unsigned int ovp_segments, reserved_segments; 3596 unsigned int main_segs, blocks_per_seg; 3597 unsigned int sit_segs, nat_segs; 3598 unsigned int sit_bitmap_size, nat_bitmap_size; 3599 unsigned int log_blocks_per_seg; 3600 unsigned int segment_count_main; 3601 unsigned int cp_pack_start_sum, cp_payload; 3602 block_t user_block_count, valid_user_blocks; 3603 block_t avail_node_count, valid_node_count; 3604 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks; 3605 int i, j; 3606 3607 total = le32_to_cpu(raw_super->segment_count); 3608 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 3609 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 3610 fsmeta += sit_segs; 3611 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 3612 fsmeta += nat_segs; 3613 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 3614 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 3615 3616 if (unlikely(fsmeta >= total)) 3617 return 1; 3618 3619 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 3620 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 3621 3622 if (!f2fs_sb_has_readonly(sbi) && 3623 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 3624 ovp_segments == 0 || reserved_segments == 0)) { 3625 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 3626 return 1; 3627 } 3628 user_block_count = le64_to_cpu(ckpt->user_block_count); 3629 segment_count_main = le32_to_cpu(raw_super->segment_count_main) + 3630 (f2fs_sb_has_readonly(sbi) ? 1 : 0); 3631 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3632 if (!user_block_count || user_block_count >= 3633 segment_count_main << log_blocks_per_seg) { 3634 f2fs_err(sbi, "Wrong user_block_count: %u", 3635 user_block_count); 3636 return 1; 3637 } 3638 3639 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 3640 if (valid_user_blocks > user_block_count) { 3641 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 3642 valid_user_blocks, user_block_count); 3643 return 1; 3644 } 3645 3646 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 3647 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 3648 if (valid_node_count > avail_node_count) { 3649 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 3650 valid_node_count, avail_node_count); 3651 return 1; 3652 } 3653 3654 main_segs = le32_to_cpu(raw_super->segment_count_main); 3655 blocks_per_seg = sbi->blocks_per_seg; 3656 3657 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3658 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 3659 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 3660 return 1; 3661 3662 if (f2fs_sb_has_readonly(sbi)) 3663 goto check_data; 3664 3665 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 3666 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3667 le32_to_cpu(ckpt->cur_node_segno[j])) { 3668 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 3669 i, j, 3670 le32_to_cpu(ckpt->cur_node_segno[i])); 3671 return 1; 3672 } 3673 } 3674 } 3675 check_data: 3676 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 3677 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 3678 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 3679 return 1; 3680 3681 if (f2fs_sb_has_readonly(sbi)) 3682 goto skip_cross; 3683 3684 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3685 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3686 le32_to_cpu(ckpt->cur_data_segno[j])) { 3687 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3688 i, j, 3689 le32_to_cpu(ckpt->cur_data_segno[i])); 3690 return 1; 3691 } 3692 } 3693 } 3694 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3695 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3696 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3697 le32_to_cpu(ckpt->cur_data_segno[j])) { 3698 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3699 i, j, 3700 le32_to_cpu(ckpt->cur_node_segno[i])); 3701 return 1; 3702 } 3703 } 3704 } 3705 skip_cross: 3706 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3707 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3708 3709 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3710 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3711 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3712 sit_bitmap_size, nat_bitmap_size); 3713 return 1; 3714 } 3715 3716 cp_pack_start_sum = __start_sum_addr(sbi); 3717 cp_payload = __cp_payload(sbi); 3718 if (cp_pack_start_sum < cp_payload + 1 || 3719 cp_pack_start_sum > blocks_per_seg - 1 - 3720 NR_CURSEG_PERSIST_TYPE) { 3721 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3722 cp_pack_start_sum); 3723 return 1; 3724 } 3725 3726 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3727 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3728 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3729 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3730 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3731 le32_to_cpu(ckpt->checksum_offset)); 3732 return 1; 3733 } 3734 3735 nat_blocks = nat_segs << log_blocks_per_seg; 3736 nat_bits_bytes = nat_blocks / BITS_PER_BYTE; 3737 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8); 3738 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) && 3739 (cp_payload + F2FS_CP_PACKS + 3740 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) { 3741 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)", 3742 cp_payload, nat_bits_blocks); 3743 return 1; 3744 } 3745 3746 if (unlikely(f2fs_cp_error(sbi))) { 3747 f2fs_err(sbi, "A bug case: need to run fsck"); 3748 return 1; 3749 } 3750 return 0; 3751 } 3752 3753 static void init_sb_info(struct f2fs_sb_info *sbi) 3754 { 3755 struct f2fs_super_block *raw_super = sbi->raw_super; 3756 int i; 3757 3758 sbi->log_sectors_per_block = 3759 le32_to_cpu(raw_super->log_sectors_per_block); 3760 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3761 sbi->blocksize = BIT(sbi->log_blocksize); 3762 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3763 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg); 3764 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3765 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3766 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3767 sbi->total_node_count = 3768 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3769 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3770 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 3771 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 3772 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 3773 sbi->cur_victim_sec = NULL_SECNO; 3774 sbi->gc_mode = GC_NORMAL; 3775 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3776 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3777 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3778 sbi->migration_granularity = sbi->segs_per_sec; 3779 sbi->seq_file_ra_mul = MIN_RA_MUL; 3780 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE; 3781 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE; 3782 spin_lock_init(&sbi->gc_remaining_trials_lock); 3783 atomic64_set(&sbi->current_atomic_write, 0); 3784 3785 sbi->dir_level = DEF_DIR_LEVEL; 3786 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3787 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3788 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3789 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3790 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3791 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3792 DEF_UMOUNT_DISCARD_TIMEOUT; 3793 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3794 3795 for (i = 0; i < NR_COUNT_TYPE; i++) 3796 atomic_set(&sbi->nr_pages[i], 0); 3797 3798 for (i = 0; i < META; i++) 3799 atomic_set(&sbi->wb_sync_req[i], 0); 3800 3801 INIT_LIST_HEAD(&sbi->s_list); 3802 mutex_init(&sbi->umount_mutex); 3803 init_f2fs_rwsem(&sbi->io_order_lock); 3804 spin_lock_init(&sbi->cp_lock); 3805 3806 sbi->dirty_device = 0; 3807 spin_lock_init(&sbi->dev_lock); 3808 3809 init_f2fs_rwsem(&sbi->sb_lock); 3810 init_f2fs_rwsem(&sbi->pin_sem); 3811 } 3812 3813 static int init_percpu_info(struct f2fs_sb_info *sbi) 3814 { 3815 int err; 3816 3817 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3818 if (err) 3819 return err; 3820 3821 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL); 3822 if (err) 3823 goto err_valid_block; 3824 3825 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3826 GFP_KERNEL); 3827 if (err) 3828 goto err_node_block; 3829 return 0; 3830 3831 err_node_block: 3832 percpu_counter_destroy(&sbi->rf_node_block_count); 3833 err_valid_block: 3834 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3835 return err; 3836 } 3837 3838 #ifdef CONFIG_BLK_DEV_ZONED 3839 3840 struct f2fs_report_zones_args { 3841 struct f2fs_sb_info *sbi; 3842 struct f2fs_dev_info *dev; 3843 }; 3844 3845 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3846 void *data) 3847 { 3848 struct f2fs_report_zones_args *rz_args = data; 3849 block_t unusable_blocks = (zone->len - zone->capacity) >> 3850 F2FS_LOG_SECTORS_PER_BLOCK; 3851 3852 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3853 return 0; 3854 3855 set_bit(idx, rz_args->dev->blkz_seq); 3856 if (!rz_args->sbi->unusable_blocks_per_sec) { 3857 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks; 3858 return 0; 3859 } 3860 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) { 3861 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n"); 3862 return -EINVAL; 3863 } 3864 return 0; 3865 } 3866 3867 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3868 { 3869 struct block_device *bdev = FDEV(devi).bdev; 3870 sector_t nr_sectors = bdev_nr_sectors(bdev); 3871 struct f2fs_report_zones_args rep_zone_arg; 3872 u64 zone_sectors; 3873 int ret; 3874 3875 if (!f2fs_sb_has_blkzoned(sbi)) 3876 return 0; 3877 3878 zone_sectors = bdev_zone_sectors(bdev); 3879 if (!is_power_of_2(zone_sectors)) { 3880 f2fs_err(sbi, "F2FS does not support non power of 2 zone sizes\n"); 3881 return -EINVAL; 3882 } 3883 3884 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3885 SECTOR_TO_BLOCK(zone_sectors)) 3886 return -EINVAL; 3887 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors); 3888 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors), 3889 sbi->blocks_per_blkz); 3890 if (nr_sectors & (zone_sectors - 1)) 3891 FDEV(devi).nr_blkz++; 3892 3893 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3894 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3895 * sizeof(unsigned long), 3896 GFP_KERNEL); 3897 if (!FDEV(devi).blkz_seq) 3898 return -ENOMEM; 3899 3900 rep_zone_arg.sbi = sbi; 3901 rep_zone_arg.dev = &FDEV(devi); 3902 3903 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3904 &rep_zone_arg); 3905 if (ret < 0) 3906 return ret; 3907 return 0; 3908 } 3909 #endif 3910 3911 /* 3912 * Read f2fs raw super block. 3913 * Because we have two copies of super block, so read both of them 3914 * to get the first valid one. If any one of them is broken, we pass 3915 * them recovery flag back to the caller. 3916 */ 3917 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3918 struct f2fs_super_block **raw_super, 3919 int *valid_super_block, int *recovery) 3920 { 3921 struct super_block *sb = sbi->sb; 3922 int block; 3923 struct buffer_head *bh; 3924 struct f2fs_super_block *super; 3925 int err = 0; 3926 3927 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3928 if (!super) 3929 return -ENOMEM; 3930 3931 for (block = 0; block < 2; block++) { 3932 bh = sb_bread(sb, block); 3933 if (!bh) { 3934 f2fs_err(sbi, "Unable to read %dth superblock", 3935 block + 1); 3936 err = -EIO; 3937 *recovery = 1; 3938 continue; 3939 } 3940 3941 /* sanity checking of raw super */ 3942 err = sanity_check_raw_super(sbi, bh); 3943 if (err) { 3944 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3945 block + 1); 3946 brelse(bh); 3947 *recovery = 1; 3948 continue; 3949 } 3950 3951 if (!*raw_super) { 3952 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3953 sizeof(*super)); 3954 *valid_super_block = block; 3955 *raw_super = super; 3956 } 3957 brelse(bh); 3958 } 3959 3960 /* No valid superblock */ 3961 if (!*raw_super) 3962 kfree(super); 3963 else 3964 err = 0; 3965 3966 return err; 3967 } 3968 3969 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3970 { 3971 struct buffer_head *bh; 3972 __u32 crc = 0; 3973 int err; 3974 3975 if ((recover && f2fs_readonly(sbi->sb)) || 3976 f2fs_hw_is_readonly(sbi)) { 3977 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3978 return -EROFS; 3979 } 3980 3981 /* we should update superblock crc here */ 3982 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3983 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3984 offsetof(struct f2fs_super_block, crc)); 3985 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3986 } 3987 3988 /* write back-up superblock first */ 3989 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3990 if (!bh) 3991 return -EIO; 3992 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3993 brelse(bh); 3994 3995 /* if we are in recovery path, skip writing valid superblock */ 3996 if (recover || err) 3997 return err; 3998 3999 /* write current valid superblock */ 4000 bh = sb_bread(sbi->sb, sbi->valid_super_block); 4001 if (!bh) 4002 return -EIO; 4003 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 4004 brelse(bh); 4005 return err; 4006 } 4007 4008 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason) 4009 { 4010 unsigned long flags; 4011 4012 spin_lock_irqsave(&sbi->error_lock, flags); 4013 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0)) 4014 sbi->stop_reason[reason]++; 4015 spin_unlock_irqrestore(&sbi->error_lock, flags); 4016 } 4017 4018 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi) 4019 { 4020 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4021 unsigned long flags; 4022 int err; 4023 4024 f2fs_down_write(&sbi->sb_lock); 4025 4026 spin_lock_irqsave(&sbi->error_lock, flags); 4027 if (sbi->error_dirty) { 4028 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors, 4029 MAX_F2FS_ERRORS); 4030 sbi->error_dirty = false; 4031 } 4032 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON); 4033 spin_unlock_irqrestore(&sbi->error_lock, flags); 4034 4035 err = f2fs_commit_super(sbi, false); 4036 4037 f2fs_up_write(&sbi->sb_lock); 4038 if (err) 4039 f2fs_err(sbi, "f2fs_commit_super fails to record err:%d", err); 4040 } 4041 4042 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag) 4043 { 4044 unsigned long flags; 4045 4046 spin_lock_irqsave(&sbi->error_lock, flags); 4047 if (!test_bit(flag, (unsigned long *)sbi->errors)) { 4048 set_bit(flag, (unsigned long *)sbi->errors); 4049 sbi->error_dirty = true; 4050 } 4051 spin_unlock_irqrestore(&sbi->error_lock, flags); 4052 } 4053 4054 static bool f2fs_update_errors(struct f2fs_sb_info *sbi) 4055 { 4056 unsigned long flags; 4057 bool need_update = false; 4058 4059 spin_lock_irqsave(&sbi->error_lock, flags); 4060 if (sbi->error_dirty) { 4061 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors, 4062 MAX_F2FS_ERRORS); 4063 sbi->error_dirty = false; 4064 need_update = true; 4065 } 4066 spin_unlock_irqrestore(&sbi->error_lock, flags); 4067 4068 return need_update; 4069 } 4070 4071 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error) 4072 { 4073 int err; 4074 4075 f2fs_down_write(&sbi->sb_lock); 4076 4077 if (!f2fs_update_errors(sbi)) 4078 goto out_unlock; 4079 4080 err = f2fs_commit_super(sbi, false); 4081 if (err) 4082 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d", 4083 error, err); 4084 out_unlock: 4085 f2fs_up_write(&sbi->sb_lock); 4086 } 4087 4088 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error) 4089 { 4090 f2fs_save_errors(sbi, error); 4091 f2fs_record_errors(sbi, error); 4092 } 4093 4094 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error) 4095 { 4096 f2fs_save_errors(sbi, error); 4097 4098 if (!sbi->error_dirty) 4099 return; 4100 if (!test_bit(error, (unsigned long *)sbi->errors)) 4101 return; 4102 schedule_work(&sbi->s_error_work); 4103 } 4104 4105 static bool system_going_down(void) 4106 { 4107 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 4108 || system_state == SYSTEM_RESTART; 4109 } 4110 4111 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason, 4112 bool irq_context) 4113 { 4114 struct super_block *sb = sbi->sb; 4115 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN; 4116 bool continue_fs = !shutdown && 4117 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE; 4118 4119 set_ckpt_flags(sbi, CP_ERROR_FLAG); 4120 4121 if (!f2fs_hw_is_readonly(sbi)) { 4122 save_stop_reason(sbi, reason); 4123 4124 if (irq_context && !shutdown) 4125 schedule_work(&sbi->s_error_work); 4126 else 4127 f2fs_record_stop_reason(sbi); 4128 } 4129 4130 /* 4131 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 4132 * could panic during 'reboot -f' as the underlying device got already 4133 * disabled. 4134 */ 4135 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC && 4136 !shutdown && !system_going_down() && 4137 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)) 4138 panic("F2FS-fs (device %s): panic forced after error\n", 4139 sb->s_id); 4140 4141 if (shutdown) 4142 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 4143 4144 /* continue filesystem operators if errors=continue */ 4145 if (continue_fs || f2fs_readonly(sb)) 4146 return; 4147 4148 f2fs_warn(sbi, "Remounting filesystem read-only"); 4149 /* 4150 * Make sure updated value of ->s_mount_flags will be visible before 4151 * ->s_flags update 4152 */ 4153 smp_wmb(); 4154 sb->s_flags |= SB_RDONLY; 4155 } 4156 4157 static void f2fs_record_error_work(struct work_struct *work) 4158 { 4159 struct f2fs_sb_info *sbi = container_of(work, 4160 struct f2fs_sb_info, s_error_work); 4161 4162 f2fs_record_stop_reason(sbi); 4163 } 4164 4165 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 4166 { 4167 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4168 unsigned int max_devices = MAX_DEVICES; 4169 unsigned int logical_blksize; 4170 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags); 4171 int i; 4172 4173 /* Initialize single device information */ 4174 if (!RDEV(0).path[0]) { 4175 if (!bdev_is_zoned(sbi->sb->s_bdev)) 4176 return 0; 4177 max_devices = 1; 4178 } 4179 4180 /* 4181 * Initialize multiple devices information, or single 4182 * zoned block device information. 4183 */ 4184 sbi->devs = f2fs_kzalloc(sbi, 4185 array_size(max_devices, 4186 sizeof(struct f2fs_dev_info)), 4187 GFP_KERNEL); 4188 if (!sbi->devs) 4189 return -ENOMEM; 4190 4191 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev); 4192 sbi->aligned_blksize = true; 4193 4194 for (i = 0; i < max_devices; i++) { 4195 if (i == 0) 4196 FDEV(0).bdev = sbi->sb->s_bdev; 4197 else if (!RDEV(i).path[0]) 4198 break; 4199 4200 if (max_devices > 1) { 4201 /* Multi-device mount */ 4202 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 4203 FDEV(i).total_segments = 4204 le32_to_cpu(RDEV(i).total_segments); 4205 if (i == 0) { 4206 FDEV(i).start_blk = 0; 4207 FDEV(i).end_blk = FDEV(i).start_blk + 4208 (FDEV(i).total_segments << 4209 sbi->log_blocks_per_seg) - 1 + 4210 le32_to_cpu(raw_super->segment0_blkaddr); 4211 } else { 4212 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 4213 FDEV(i).end_blk = FDEV(i).start_blk + 4214 (FDEV(i).total_segments << 4215 sbi->log_blocks_per_seg) - 1; 4216 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 4217 mode, sbi->sb->s_type, NULL); 4218 } 4219 } 4220 if (IS_ERR(FDEV(i).bdev)) 4221 return PTR_ERR(FDEV(i).bdev); 4222 4223 /* to release errored devices */ 4224 sbi->s_ndevs = i + 1; 4225 4226 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev)) 4227 sbi->aligned_blksize = false; 4228 4229 #ifdef CONFIG_BLK_DEV_ZONED 4230 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 4231 !f2fs_sb_has_blkzoned(sbi)) { 4232 f2fs_err(sbi, "Zoned block device feature not enabled"); 4233 return -EINVAL; 4234 } 4235 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 4236 if (init_blkz_info(sbi, i)) { 4237 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 4238 return -EINVAL; 4239 } 4240 if (max_devices == 1) 4241 break; 4242 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 4243 i, FDEV(i).path, 4244 FDEV(i).total_segments, 4245 FDEV(i).start_blk, FDEV(i).end_blk, 4246 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 4247 "Host-aware" : "Host-managed"); 4248 continue; 4249 } 4250 #endif 4251 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 4252 i, FDEV(i).path, 4253 FDEV(i).total_segments, 4254 FDEV(i).start_blk, FDEV(i).end_blk); 4255 } 4256 f2fs_info(sbi, 4257 "IO Block Size: %8ld KB", F2FS_IO_SIZE_KB(sbi)); 4258 return 0; 4259 } 4260 4261 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 4262 { 4263 #if IS_ENABLED(CONFIG_UNICODE) 4264 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 4265 const struct f2fs_sb_encodings *encoding_info; 4266 struct unicode_map *encoding; 4267 __u16 encoding_flags; 4268 4269 encoding_info = f2fs_sb_read_encoding(sbi->raw_super); 4270 if (!encoding_info) { 4271 f2fs_err(sbi, 4272 "Encoding requested by superblock is unknown"); 4273 return -EINVAL; 4274 } 4275 4276 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags); 4277 encoding = utf8_load(encoding_info->version); 4278 if (IS_ERR(encoding)) { 4279 f2fs_err(sbi, 4280 "can't mount with superblock charset: %s-%u.%u.%u " 4281 "not supported by the kernel. flags: 0x%x.", 4282 encoding_info->name, 4283 unicode_major(encoding_info->version), 4284 unicode_minor(encoding_info->version), 4285 unicode_rev(encoding_info->version), 4286 encoding_flags); 4287 return PTR_ERR(encoding); 4288 } 4289 f2fs_info(sbi, "Using encoding defined by superblock: " 4290 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 4291 unicode_major(encoding_info->version), 4292 unicode_minor(encoding_info->version), 4293 unicode_rev(encoding_info->version), 4294 encoding_flags); 4295 4296 sbi->sb->s_encoding = encoding; 4297 sbi->sb->s_encoding_flags = encoding_flags; 4298 } 4299 #else 4300 if (f2fs_sb_has_casefold(sbi)) { 4301 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 4302 return -EINVAL; 4303 } 4304 #endif 4305 return 0; 4306 } 4307 4308 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 4309 { 4310 /* adjust parameters according to the volume size */ 4311 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) { 4312 if (f2fs_block_unit_discard(sbi)) 4313 SM_I(sbi)->dcc_info->discard_granularity = 4314 MIN_DISCARD_GRANULARITY; 4315 if (!f2fs_lfs_mode(sbi)) 4316 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) | 4317 BIT(F2FS_IPU_HONOR_OPU_WRITE); 4318 } 4319 4320 sbi->readdir_ra = true; 4321 } 4322 4323 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 4324 { 4325 struct f2fs_sb_info *sbi; 4326 struct f2fs_super_block *raw_super; 4327 struct inode *root; 4328 int err; 4329 bool skip_recovery = false, need_fsck = false; 4330 char *options = NULL; 4331 int recovery, i, valid_super_block; 4332 struct curseg_info *seg_i; 4333 int retry_cnt = 1; 4334 #ifdef CONFIG_QUOTA 4335 bool quota_enabled = false; 4336 #endif 4337 4338 try_onemore: 4339 err = -EINVAL; 4340 raw_super = NULL; 4341 valid_super_block = -1; 4342 recovery = 0; 4343 4344 /* allocate memory for f2fs-specific super block info */ 4345 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 4346 if (!sbi) 4347 return -ENOMEM; 4348 4349 sbi->sb = sb; 4350 4351 /* initialize locks within allocated memory */ 4352 init_f2fs_rwsem(&sbi->gc_lock); 4353 mutex_init(&sbi->writepages); 4354 init_f2fs_rwsem(&sbi->cp_global_sem); 4355 init_f2fs_rwsem(&sbi->node_write); 4356 init_f2fs_rwsem(&sbi->node_change); 4357 spin_lock_init(&sbi->stat_lock); 4358 init_f2fs_rwsem(&sbi->cp_rwsem); 4359 init_f2fs_rwsem(&sbi->quota_sem); 4360 init_waitqueue_head(&sbi->cp_wait); 4361 spin_lock_init(&sbi->error_lock); 4362 4363 for (i = 0; i < NR_INODE_TYPE; i++) { 4364 INIT_LIST_HEAD(&sbi->inode_list[i]); 4365 spin_lock_init(&sbi->inode_lock[i]); 4366 } 4367 mutex_init(&sbi->flush_lock); 4368 4369 /* Load the checksum driver */ 4370 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 4371 if (IS_ERR(sbi->s_chksum_driver)) { 4372 f2fs_err(sbi, "Cannot load crc32 driver."); 4373 err = PTR_ERR(sbi->s_chksum_driver); 4374 sbi->s_chksum_driver = NULL; 4375 goto free_sbi; 4376 } 4377 4378 /* set a block size */ 4379 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 4380 f2fs_err(sbi, "unable to set blocksize"); 4381 goto free_sbi; 4382 } 4383 4384 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 4385 &recovery); 4386 if (err) 4387 goto free_sbi; 4388 4389 sb->s_fs_info = sbi; 4390 sbi->raw_super = raw_super; 4391 4392 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work); 4393 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS); 4394 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON); 4395 4396 /* precompute checksum seed for metadata */ 4397 if (f2fs_sb_has_inode_chksum(sbi)) 4398 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 4399 sizeof(raw_super->uuid)); 4400 4401 default_options(sbi, false); 4402 /* parse mount options */ 4403 options = kstrdup((const char *)data, GFP_KERNEL); 4404 if (data && !options) { 4405 err = -ENOMEM; 4406 goto free_sb_buf; 4407 } 4408 4409 err = parse_options(sb, options, false); 4410 if (err) 4411 goto free_options; 4412 4413 sb->s_maxbytes = max_file_blocks(NULL) << 4414 le32_to_cpu(raw_super->log_blocksize); 4415 sb->s_max_links = F2FS_LINK_MAX; 4416 4417 err = f2fs_setup_casefold(sbi); 4418 if (err) 4419 goto free_options; 4420 4421 #ifdef CONFIG_QUOTA 4422 sb->dq_op = &f2fs_quota_operations; 4423 sb->s_qcop = &f2fs_quotactl_ops; 4424 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4425 4426 if (f2fs_sb_has_quota_ino(sbi)) { 4427 for (i = 0; i < MAXQUOTAS; i++) { 4428 if (f2fs_qf_ino(sbi->sb, i)) 4429 sbi->nquota_files++; 4430 } 4431 } 4432 #endif 4433 4434 sb->s_op = &f2fs_sops; 4435 #ifdef CONFIG_FS_ENCRYPTION 4436 sb->s_cop = &f2fs_cryptops; 4437 #endif 4438 #ifdef CONFIG_FS_VERITY 4439 sb->s_vop = &f2fs_verityops; 4440 #endif 4441 sb->s_xattr = f2fs_xattr_handlers; 4442 sb->s_export_op = &f2fs_export_ops; 4443 sb->s_magic = F2FS_SUPER_MAGIC; 4444 sb->s_time_gran = 1; 4445 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4446 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 4447 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 4448 sb->s_iflags |= SB_I_CGROUPWB; 4449 4450 /* init f2fs-specific super block info */ 4451 sbi->valid_super_block = valid_super_block; 4452 4453 /* disallow all the data/node/meta page writes */ 4454 set_sbi_flag(sbi, SBI_POR_DOING); 4455 4456 err = f2fs_init_write_merge_io(sbi); 4457 if (err) 4458 goto free_bio_info; 4459 4460 init_sb_info(sbi); 4461 4462 err = f2fs_init_iostat(sbi); 4463 if (err) 4464 goto free_bio_info; 4465 4466 err = init_percpu_info(sbi); 4467 if (err) 4468 goto free_iostat; 4469 4470 if (F2FS_IO_ALIGNED(sbi)) { 4471 sbi->write_io_dummy = 4472 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 4473 if (!sbi->write_io_dummy) { 4474 err = -ENOMEM; 4475 goto free_percpu; 4476 } 4477 } 4478 4479 /* init per sbi slab cache */ 4480 err = f2fs_init_xattr_caches(sbi); 4481 if (err) 4482 goto free_io_dummy; 4483 err = f2fs_init_page_array_cache(sbi); 4484 if (err) 4485 goto free_xattr_cache; 4486 4487 /* get an inode for meta space */ 4488 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 4489 if (IS_ERR(sbi->meta_inode)) { 4490 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 4491 err = PTR_ERR(sbi->meta_inode); 4492 goto free_page_array_cache; 4493 } 4494 4495 err = f2fs_get_valid_checkpoint(sbi); 4496 if (err) { 4497 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 4498 goto free_meta_inode; 4499 } 4500 4501 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 4502 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 4503 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 4504 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4505 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 4506 } 4507 4508 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 4509 set_sbi_flag(sbi, SBI_NEED_FSCK); 4510 4511 /* Initialize device list */ 4512 err = f2fs_scan_devices(sbi); 4513 if (err) { 4514 f2fs_err(sbi, "Failed to find devices"); 4515 goto free_devices; 4516 } 4517 4518 err = f2fs_init_post_read_wq(sbi); 4519 if (err) { 4520 f2fs_err(sbi, "Failed to initialize post read workqueue"); 4521 goto free_devices; 4522 } 4523 4524 sbi->total_valid_node_count = 4525 le32_to_cpu(sbi->ckpt->valid_node_count); 4526 percpu_counter_set(&sbi->total_valid_inode_count, 4527 le32_to_cpu(sbi->ckpt->valid_inode_count)); 4528 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 4529 sbi->total_valid_block_count = 4530 le64_to_cpu(sbi->ckpt->valid_block_count); 4531 sbi->last_valid_block_count = sbi->total_valid_block_count; 4532 sbi->reserved_blocks = 0; 4533 sbi->current_reserved_blocks = 0; 4534 limit_reserve_root(sbi); 4535 adjust_unusable_cap_perc(sbi); 4536 4537 f2fs_init_extent_cache_info(sbi); 4538 4539 f2fs_init_ino_entry_info(sbi); 4540 4541 f2fs_init_fsync_node_info(sbi); 4542 4543 /* setup checkpoint request control and start checkpoint issue thread */ 4544 f2fs_init_ckpt_req_control(sbi); 4545 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) && 4546 test_opt(sbi, MERGE_CHECKPOINT)) { 4547 err = f2fs_start_ckpt_thread(sbi); 4548 if (err) { 4549 f2fs_err(sbi, 4550 "Failed to start F2FS issue_checkpoint_thread (%d)", 4551 err); 4552 goto stop_ckpt_thread; 4553 } 4554 } 4555 4556 /* setup f2fs internal modules */ 4557 err = f2fs_build_segment_manager(sbi); 4558 if (err) { 4559 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 4560 err); 4561 goto free_sm; 4562 } 4563 err = f2fs_build_node_manager(sbi); 4564 if (err) { 4565 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 4566 err); 4567 goto free_nm; 4568 } 4569 4570 err = adjust_reserved_segment(sbi); 4571 if (err) 4572 goto free_nm; 4573 4574 /* For write statistics */ 4575 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 4576 4577 /* Read accumulated write IO statistics if exists */ 4578 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 4579 if (__exist_node_summaries(sbi)) 4580 sbi->kbytes_written = 4581 le64_to_cpu(seg_i->journal->info.kbytes_written); 4582 4583 f2fs_build_gc_manager(sbi); 4584 4585 err = f2fs_build_stats(sbi); 4586 if (err) 4587 goto free_nm; 4588 4589 /* get an inode for node space */ 4590 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 4591 if (IS_ERR(sbi->node_inode)) { 4592 f2fs_err(sbi, "Failed to read node inode"); 4593 err = PTR_ERR(sbi->node_inode); 4594 goto free_stats; 4595 } 4596 4597 /* read root inode and dentry */ 4598 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 4599 if (IS_ERR(root)) { 4600 f2fs_err(sbi, "Failed to read root inode"); 4601 err = PTR_ERR(root); 4602 goto free_node_inode; 4603 } 4604 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 4605 !root->i_size || !root->i_nlink) { 4606 iput(root); 4607 err = -EINVAL; 4608 goto free_node_inode; 4609 } 4610 4611 sb->s_root = d_make_root(root); /* allocate root dentry */ 4612 if (!sb->s_root) { 4613 err = -ENOMEM; 4614 goto free_node_inode; 4615 } 4616 4617 err = f2fs_init_compress_inode(sbi); 4618 if (err) 4619 goto free_root_inode; 4620 4621 err = f2fs_register_sysfs(sbi); 4622 if (err) 4623 goto free_compress_inode; 4624 4625 #ifdef CONFIG_QUOTA 4626 /* Enable quota usage during mount */ 4627 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 4628 err = f2fs_enable_quotas(sb); 4629 if (err) 4630 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 4631 } 4632 4633 quota_enabled = f2fs_recover_quota_begin(sbi); 4634 #endif 4635 /* if there are any orphan inodes, free them */ 4636 err = f2fs_recover_orphan_inodes(sbi); 4637 if (err) 4638 goto free_meta; 4639 4640 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 4641 goto reset_checkpoint; 4642 4643 /* recover fsynced data */ 4644 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 4645 !test_opt(sbi, NORECOVERY)) { 4646 /* 4647 * mount should be failed, when device has readonly mode, and 4648 * previous checkpoint was not done by clean system shutdown. 4649 */ 4650 if (f2fs_hw_is_readonly(sbi)) { 4651 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4652 err = f2fs_recover_fsync_data(sbi, true); 4653 if (err > 0) { 4654 err = -EROFS; 4655 f2fs_err(sbi, "Need to recover fsync data, but " 4656 "write access unavailable, please try " 4657 "mount w/ disable_roll_forward or norecovery"); 4658 } 4659 if (err < 0) 4660 goto free_meta; 4661 } 4662 f2fs_info(sbi, "write access unavailable, skipping recovery"); 4663 goto reset_checkpoint; 4664 } 4665 4666 if (need_fsck) 4667 set_sbi_flag(sbi, SBI_NEED_FSCK); 4668 4669 if (skip_recovery) 4670 goto reset_checkpoint; 4671 4672 err = f2fs_recover_fsync_data(sbi, false); 4673 if (err < 0) { 4674 if (err != -ENOMEM) 4675 skip_recovery = true; 4676 need_fsck = true; 4677 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 4678 err); 4679 goto free_meta; 4680 } 4681 } else { 4682 err = f2fs_recover_fsync_data(sbi, true); 4683 4684 if (!f2fs_readonly(sb) && err > 0) { 4685 err = -EINVAL; 4686 f2fs_err(sbi, "Need to recover fsync data"); 4687 goto free_meta; 4688 } 4689 } 4690 4691 #ifdef CONFIG_QUOTA 4692 f2fs_recover_quota_end(sbi, quota_enabled); 4693 #endif 4694 4695 /* 4696 * If the f2fs is not readonly and fsync data recovery succeeds, 4697 * check zoned block devices' write pointer consistency. 4698 */ 4699 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 4700 err = f2fs_check_write_pointer(sbi); 4701 if (err) 4702 goto free_meta; 4703 } 4704 4705 reset_checkpoint: 4706 f2fs_init_inmem_curseg(sbi); 4707 4708 /* f2fs_recover_fsync_data() cleared this already */ 4709 clear_sbi_flag(sbi, SBI_POR_DOING); 4710 4711 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 4712 err = f2fs_disable_checkpoint(sbi); 4713 if (err) 4714 goto sync_free_meta; 4715 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 4716 f2fs_enable_checkpoint(sbi); 4717 } 4718 4719 /* 4720 * If filesystem is not mounted as read-only then 4721 * do start the gc_thread. 4722 */ 4723 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF || 4724 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) { 4725 /* After POR, we can run background GC thread.*/ 4726 err = f2fs_start_gc_thread(sbi); 4727 if (err) 4728 goto sync_free_meta; 4729 } 4730 kvfree(options); 4731 4732 /* recover broken superblock */ 4733 if (recovery) { 4734 err = f2fs_commit_super(sbi, true); 4735 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 4736 sbi->valid_super_block ? 1 : 2, err); 4737 } 4738 4739 f2fs_join_shrinker(sbi); 4740 4741 f2fs_tuning_parameters(sbi); 4742 4743 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 4744 cur_cp_version(F2FS_CKPT(sbi))); 4745 f2fs_update_time(sbi, CP_TIME); 4746 f2fs_update_time(sbi, REQ_TIME); 4747 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4748 return 0; 4749 4750 sync_free_meta: 4751 /* safe to flush all the data */ 4752 sync_filesystem(sbi->sb); 4753 retry_cnt = 0; 4754 4755 free_meta: 4756 #ifdef CONFIG_QUOTA 4757 f2fs_truncate_quota_inode_pages(sb); 4758 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 4759 f2fs_quota_off_umount(sbi->sb); 4760 #endif 4761 /* 4762 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 4763 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 4764 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 4765 * falls into an infinite loop in f2fs_sync_meta_pages(). 4766 */ 4767 truncate_inode_pages_final(META_MAPPING(sbi)); 4768 /* evict some inodes being cached by GC */ 4769 evict_inodes(sb); 4770 f2fs_unregister_sysfs(sbi); 4771 free_compress_inode: 4772 f2fs_destroy_compress_inode(sbi); 4773 free_root_inode: 4774 dput(sb->s_root); 4775 sb->s_root = NULL; 4776 free_node_inode: 4777 f2fs_release_ino_entry(sbi, true); 4778 truncate_inode_pages_final(NODE_MAPPING(sbi)); 4779 iput(sbi->node_inode); 4780 sbi->node_inode = NULL; 4781 free_stats: 4782 f2fs_destroy_stats(sbi); 4783 free_nm: 4784 /* stop discard thread before destroying node manager */ 4785 f2fs_stop_discard_thread(sbi); 4786 f2fs_destroy_node_manager(sbi); 4787 free_sm: 4788 f2fs_destroy_segment_manager(sbi); 4789 stop_ckpt_thread: 4790 f2fs_stop_ckpt_thread(sbi); 4791 /* flush s_error_work before sbi destroy */ 4792 flush_work(&sbi->s_error_work); 4793 f2fs_destroy_post_read_wq(sbi); 4794 free_devices: 4795 destroy_device_list(sbi); 4796 kvfree(sbi->ckpt); 4797 free_meta_inode: 4798 make_bad_inode(sbi->meta_inode); 4799 iput(sbi->meta_inode); 4800 sbi->meta_inode = NULL; 4801 free_page_array_cache: 4802 f2fs_destroy_page_array_cache(sbi); 4803 free_xattr_cache: 4804 f2fs_destroy_xattr_caches(sbi); 4805 free_io_dummy: 4806 mempool_destroy(sbi->write_io_dummy); 4807 free_percpu: 4808 destroy_percpu_info(sbi); 4809 free_iostat: 4810 f2fs_destroy_iostat(sbi); 4811 free_bio_info: 4812 for (i = 0; i < NR_PAGE_TYPE; i++) 4813 kvfree(sbi->write_io[i]); 4814 4815 #if IS_ENABLED(CONFIG_UNICODE) 4816 utf8_unload(sb->s_encoding); 4817 sb->s_encoding = NULL; 4818 #endif 4819 free_options: 4820 #ifdef CONFIG_QUOTA 4821 for (i = 0; i < MAXQUOTAS; i++) 4822 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 4823 #endif 4824 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 4825 kvfree(options); 4826 free_sb_buf: 4827 kfree(raw_super); 4828 free_sbi: 4829 if (sbi->s_chksum_driver) 4830 crypto_free_shash(sbi->s_chksum_driver); 4831 kfree(sbi); 4832 4833 /* give only one another chance */ 4834 if (retry_cnt > 0 && skip_recovery) { 4835 retry_cnt--; 4836 shrink_dcache_sb(sb); 4837 goto try_onemore; 4838 } 4839 return err; 4840 } 4841 4842 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 4843 const char *dev_name, void *data) 4844 { 4845 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 4846 } 4847 4848 static void kill_f2fs_super(struct super_block *sb) 4849 { 4850 if (sb->s_root) { 4851 struct f2fs_sb_info *sbi = F2FS_SB(sb); 4852 4853 set_sbi_flag(sbi, SBI_IS_CLOSE); 4854 f2fs_stop_gc_thread(sbi); 4855 f2fs_stop_discard_thread(sbi); 4856 4857 #ifdef CONFIG_F2FS_FS_COMPRESSION 4858 /* 4859 * latter evict_inode() can bypass checking and invalidating 4860 * compress inode cache. 4861 */ 4862 if (test_opt(sbi, COMPRESS_CACHE)) 4863 truncate_inode_pages_final(COMPRESS_MAPPING(sbi)); 4864 #endif 4865 4866 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 4867 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4868 struct cp_control cpc = { 4869 .reason = CP_UMOUNT, 4870 }; 4871 f2fs_write_checkpoint(sbi, &cpc); 4872 } 4873 4874 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 4875 sb->s_flags &= ~SB_RDONLY; 4876 } 4877 kill_block_super(sb); 4878 } 4879 4880 static struct file_system_type f2fs_fs_type = { 4881 .owner = THIS_MODULE, 4882 .name = "f2fs", 4883 .mount = f2fs_mount, 4884 .kill_sb = kill_f2fs_super, 4885 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 4886 }; 4887 MODULE_ALIAS_FS("f2fs"); 4888 4889 static int __init init_inodecache(void) 4890 { 4891 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 4892 sizeof(struct f2fs_inode_info), 0, 4893 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 4894 return f2fs_inode_cachep ? 0 : -ENOMEM; 4895 } 4896 4897 static void destroy_inodecache(void) 4898 { 4899 /* 4900 * Make sure all delayed rcu free inodes are flushed before we 4901 * destroy cache. 4902 */ 4903 rcu_barrier(); 4904 kmem_cache_destroy(f2fs_inode_cachep); 4905 } 4906 4907 static int __init init_f2fs_fs(void) 4908 { 4909 int err; 4910 4911 if (PAGE_SIZE != F2FS_BLKSIZE) { 4912 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 4913 PAGE_SIZE, F2FS_BLKSIZE); 4914 return -EINVAL; 4915 } 4916 4917 err = init_inodecache(); 4918 if (err) 4919 goto fail; 4920 err = f2fs_create_node_manager_caches(); 4921 if (err) 4922 goto free_inodecache; 4923 err = f2fs_create_segment_manager_caches(); 4924 if (err) 4925 goto free_node_manager_caches; 4926 err = f2fs_create_checkpoint_caches(); 4927 if (err) 4928 goto free_segment_manager_caches; 4929 err = f2fs_create_recovery_cache(); 4930 if (err) 4931 goto free_checkpoint_caches; 4932 err = f2fs_create_extent_cache(); 4933 if (err) 4934 goto free_recovery_cache; 4935 err = f2fs_create_garbage_collection_cache(); 4936 if (err) 4937 goto free_extent_cache; 4938 err = f2fs_init_sysfs(); 4939 if (err) 4940 goto free_garbage_collection_cache; 4941 err = register_shrinker(&f2fs_shrinker_info, "f2fs-shrinker"); 4942 if (err) 4943 goto free_sysfs; 4944 err = register_filesystem(&f2fs_fs_type); 4945 if (err) 4946 goto free_shrinker; 4947 f2fs_create_root_stats(); 4948 err = f2fs_init_post_read_processing(); 4949 if (err) 4950 goto free_root_stats; 4951 err = f2fs_init_iostat_processing(); 4952 if (err) 4953 goto free_post_read; 4954 err = f2fs_init_bio_entry_cache(); 4955 if (err) 4956 goto free_iostat; 4957 err = f2fs_init_bioset(); 4958 if (err) 4959 goto free_bio_entry_cache; 4960 err = f2fs_init_compress_mempool(); 4961 if (err) 4962 goto free_bioset; 4963 err = f2fs_init_compress_cache(); 4964 if (err) 4965 goto free_compress_mempool; 4966 err = f2fs_create_casefold_cache(); 4967 if (err) 4968 goto free_compress_cache; 4969 return 0; 4970 free_compress_cache: 4971 f2fs_destroy_compress_cache(); 4972 free_compress_mempool: 4973 f2fs_destroy_compress_mempool(); 4974 free_bioset: 4975 f2fs_destroy_bioset(); 4976 free_bio_entry_cache: 4977 f2fs_destroy_bio_entry_cache(); 4978 free_iostat: 4979 f2fs_destroy_iostat_processing(); 4980 free_post_read: 4981 f2fs_destroy_post_read_processing(); 4982 free_root_stats: 4983 f2fs_destroy_root_stats(); 4984 unregister_filesystem(&f2fs_fs_type); 4985 free_shrinker: 4986 unregister_shrinker(&f2fs_shrinker_info); 4987 free_sysfs: 4988 f2fs_exit_sysfs(); 4989 free_garbage_collection_cache: 4990 f2fs_destroy_garbage_collection_cache(); 4991 free_extent_cache: 4992 f2fs_destroy_extent_cache(); 4993 free_recovery_cache: 4994 f2fs_destroy_recovery_cache(); 4995 free_checkpoint_caches: 4996 f2fs_destroy_checkpoint_caches(); 4997 free_segment_manager_caches: 4998 f2fs_destroy_segment_manager_caches(); 4999 free_node_manager_caches: 5000 f2fs_destroy_node_manager_caches(); 5001 free_inodecache: 5002 destroy_inodecache(); 5003 fail: 5004 return err; 5005 } 5006 5007 static void __exit exit_f2fs_fs(void) 5008 { 5009 f2fs_destroy_casefold_cache(); 5010 f2fs_destroy_compress_cache(); 5011 f2fs_destroy_compress_mempool(); 5012 f2fs_destroy_bioset(); 5013 f2fs_destroy_bio_entry_cache(); 5014 f2fs_destroy_iostat_processing(); 5015 f2fs_destroy_post_read_processing(); 5016 f2fs_destroy_root_stats(); 5017 unregister_filesystem(&f2fs_fs_type); 5018 unregister_shrinker(&f2fs_shrinker_info); 5019 f2fs_exit_sysfs(); 5020 f2fs_destroy_garbage_collection_cache(); 5021 f2fs_destroy_extent_cache(); 5022 f2fs_destroy_recovery_cache(); 5023 f2fs_destroy_checkpoint_caches(); 5024 f2fs_destroy_segment_manager_caches(); 5025 f2fs_destroy_node_manager_caches(); 5026 destroy_inodecache(); 5027 } 5028 5029 module_init(init_f2fs_fs) 5030 module_exit(exit_f2fs_fs) 5031 5032 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 5033 MODULE_DESCRIPTION("Flash Friendly File System"); 5034 MODULE_LICENSE("GPL"); 5035 MODULE_SOFTDEP("pre: crc32"); 5036 5037